xref: /freebsd/sys/dev/e1000/if_em.c (revision 0957b409)
1 /*-
2  * SPDX-License-Identifier: BSD-2-Clause
3  *
4  * Copyright (c) 2016 Nicole Graziano <nicole@nextbsd.org>
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  *
16  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
17  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
20  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
26  * SUCH DAMAGE.
27  */
28 
29 /* $FreeBSD$ */
30 #include "if_em.h"
31 #include <sys/sbuf.h>
32 #include <machine/_inttypes.h>
33 
34 #define em_mac_min e1000_82547
35 #define igb_mac_min e1000_82575
36 
37 /*********************************************************************
38  *  Driver version:
39  *********************************************************************/
40 char em_driver_version[] = "7.6.1-k";
41 
42 /*********************************************************************
43  *  PCI Device ID Table
44  *
45  *  Used by probe to select devices to load on
46  *  Last field stores an index into e1000_strings
47  *  Last entry must be all 0s
48  *
49  *  { Vendor ID, Device ID, SubVendor ID, SubDevice ID, String Index }
50  *********************************************************************/
51 
52 static pci_vendor_info_t em_vendor_info_array[] =
53 {
54 	/* Intel(R) PRO/1000 Network Connection - Legacy em*/
55 	PVID(0x8086, E1000_DEV_ID_82540EM, "Intel(R) PRO/1000 Network Connection"),
56 	PVID(0x8086, E1000_DEV_ID_82540EM_LOM, "Intel(R) PRO/1000 Network Connection"),
57 	PVID(0x8086, E1000_DEV_ID_82540EP, "Intel(R) PRO/1000 Network Connection"),
58 	PVID(0x8086, E1000_DEV_ID_82540EP_LOM, "Intel(R) PRO/1000 Network Connection"),
59 	PVID(0x8086, E1000_DEV_ID_82540EP_LP, "Intel(R) PRO/1000 Network Connection"),
60 
61 	PVID(0x8086, E1000_DEV_ID_82541EI, "Intel(R) PRO/1000 Network Connection"),
62 	PVID(0x8086, E1000_DEV_ID_82541ER, "Intel(R) PRO/1000 Network Connection"),
63 	PVID(0x8086, E1000_DEV_ID_82541ER_LOM, "Intel(R) PRO/1000 Network Connection"),
64 	PVID(0x8086, E1000_DEV_ID_82541EI_MOBILE, "Intel(R) PRO/1000 Network Connection"),
65 	PVID(0x8086, E1000_DEV_ID_82541GI, "Intel(R) PRO/1000 Network Connection"),
66 	PVID(0x8086, E1000_DEV_ID_82541GI_LF, "Intel(R) PRO/1000 Network Connection"),
67 	PVID(0x8086, E1000_DEV_ID_82541GI_MOBILE, "Intel(R) PRO/1000 Network Connection"),
68 
69 	PVID(0x8086, E1000_DEV_ID_82542, "Intel(R) PRO/1000 Network Connection"),
70 
71 	PVID(0x8086, E1000_DEV_ID_82543GC_FIBER, "Intel(R) PRO/1000 Network Connection"),
72 	PVID(0x8086, E1000_DEV_ID_82543GC_COPPER, "Intel(R) PRO/1000 Network Connection"),
73 
74 	PVID(0x8086, E1000_DEV_ID_82544EI_COPPER, "Intel(R) PRO/1000 Network Connection"),
75 	PVID(0x8086, E1000_DEV_ID_82544EI_FIBER, "Intel(R) PRO/1000 Network Connection"),
76 	PVID(0x8086, E1000_DEV_ID_82544GC_COPPER, "Intel(R) PRO/1000 Network Connection"),
77 	PVID(0x8086, E1000_DEV_ID_82544GC_LOM, "Intel(R) PRO/1000 Network Connection"),
78 
79 	PVID(0x8086, E1000_DEV_ID_82545EM_COPPER, "Intel(R) PRO/1000 Network Connection"),
80 	PVID(0x8086, E1000_DEV_ID_82545EM_FIBER, "Intel(R) PRO/1000 Network Connection"),
81 	PVID(0x8086, E1000_DEV_ID_82545GM_COPPER, "Intel(R) PRO/1000 Network Connection"),
82 	PVID(0x8086, E1000_DEV_ID_82545GM_FIBER, "Intel(R) PRO/1000 Network Connection"),
83 	PVID(0x8086, E1000_DEV_ID_82545GM_SERDES, "Intel(R) PRO/1000 Network Connection"),
84 
85 	PVID(0x8086, E1000_DEV_ID_82546EB_COPPER, "Intel(R) PRO/1000 Network Connection"),
86 	PVID(0x8086, E1000_DEV_ID_82546EB_FIBER, "Intel(R) PRO/1000 Network Connection"),
87 	PVID(0x8086, E1000_DEV_ID_82546EB_QUAD_COPPER, "Intel(R) PRO/1000 Network Connection"),
88 	PVID(0x8086, E1000_DEV_ID_82546GB_COPPER, "Intel(R) PRO/1000 Network Connection"),
89 	PVID(0x8086, E1000_DEV_ID_82546GB_FIBER, "Intel(R) PRO/1000 Network Connection"),
90 	PVID(0x8086, E1000_DEV_ID_82546GB_SERDES, "Intel(R) PRO/1000 Network Connection"),
91 	PVID(0x8086, E1000_DEV_ID_82546GB_PCIE, "Intel(R) PRO/1000 Network Connection"),
92 	PVID(0x8086, E1000_DEV_ID_82546GB_QUAD_COPPER, "Intel(R) PRO/1000 Network Connection"),
93 	PVID(0x8086, E1000_DEV_ID_82546GB_QUAD_COPPER_KSP3, "Intel(R) PRO/1000 Network Connection"),
94 
95 	PVID(0x8086, E1000_DEV_ID_82547EI, "Intel(R) PRO/1000 Network Connection"),
96 	PVID(0x8086, E1000_DEV_ID_82547EI_MOBILE, "Intel(R) PRO/1000 Network Connection"),
97 	PVID(0x8086, E1000_DEV_ID_82547GI, "Intel(R) PRO/1000 Network Connection"),
98 
99 	/* Intel(R) PRO/1000 Network Connection - em */
100 	PVID(0x8086, E1000_DEV_ID_82571EB_COPPER, "Intel(R) PRO/1000 Network Connection"),
101 	PVID(0x8086, E1000_DEV_ID_82571EB_FIBER, "Intel(R) PRO/1000 Network Connection"),
102 	PVID(0x8086, E1000_DEV_ID_82571EB_SERDES, "Intel(R) PRO/1000 Network Connection"),
103 	PVID(0x8086, E1000_DEV_ID_82571EB_SERDES_DUAL, "Intel(R) PRO/1000 Network Connection"),
104 	PVID(0x8086, E1000_DEV_ID_82571EB_SERDES_QUAD, "Intel(R) PRO/1000 Network Connection"),
105 	PVID(0x8086, E1000_DEV_ID_82571EB_QUAD_COPPER, "Intel(R) PRO/1000 Network Connection"),
106 	PVID(0x8086, E1000_DEV_ID_82571EB_QUAD_COPPER_LP, "Intel(R) PRO/1000 Network Connection"),
107 	PVID(0x8086, E1000_DEV_ID_82571EB_QUAD_FIBER, "Intel(R) PRO/1000 Network Connection"),
108 	PVID(0x8086, E1000_DEV_ID_82571PT_QUAD_COPPER, "Intel(R) PRO/1000 Network Connection"),
109 	PVID(0x8086, E1000_DEV_ID_82572EI, "Intel(R) PRO/1000 Network Connection"),
110 	PVID(0x8086, E1000_DEV_ID_82572EI_COPPER, "Intel(R) PRO/1000 Network Connection"),
111 	PVID(0x8086, E1000_DEV_ID_82572EI_FIBER, "Intel(R) PRO/1000 Network Connection"),
112 	PVID(0x8086, E1000_DEV_ID_82572EI_SERDES, "Intel(R) PRO/1000 Network Connection"),
113 	PVID(0x8086, E1000_DEV_ID_82573E, "Intel(R) PRO/1000 Network Connection"),
114 	PVID(0x8086, E1000_DEV_ID_82573E_IAMT, "Intel(R) PRO/1000 Network Connection"),
115 	PVID(0x8086, E1000_DEV_ID_82573L, "Intel(R) PRO/1000 Network Connection"),
116 	PVID(0x8086, E1000_DEV_ID_82583V, "Intel(R) PRO/1000 Network Connection"),
117 	PVID(0x8086, E1000_DEV_ID_80003ES2LAN_COPPER_SPT, "Intel(R) PRO/1000 Network Connection"),
118 	PVID(0x8086, E1000_DEV_ID_80003ES2LAN_SERDES_SPT, "Intel(R) PRO/1000 Network Connection"),
119 	PVID(0x8086, E1000_DEV_ID_80003ES2LAN_COPPER_DPT, "Intel(R) PRO/1000 Network Connection"),
120 	PVID(0x8086, E1000_DEV_ID_80003ES2LAN_SERDES_DPT, "Intel(R) PRO/1000 Network Connection"),
121 	PVID(0x8086, E1000_DEV_ID_ICH8_IGP_M_AMT, "Intel(R) PRO/1000 Network Connection"),
122 	PVID(0x8086, E1000_DEV_ID_ICH8_IGP_AMT, "Intel(R) PRO/1000 Network Connection"),
123 	PVID(0x8086, E1000_DEV_ID_ICH8_IGP_C, "Intel(R) PRO/1000 Network Connection"),
124 	PVID(0x8086, E1000_DEV_ID_ICH8_IFE, "Intel(R) PRO/1000 Network Connection"),
125 	PVID(0x8086, E1000_DEV_ID_ICH8_IFE_GT, "Intel(R) PRO/1000 Network Connection"),
126 	PVID(0x8086, E1000_DEV_ID_ICH8_IFE_G, "Intel(R) PRO/1000 Network Connection"),
127 	PVID(0x8086, E1000_DEV_ID_ICH8_IGP_M, "Intel(R) PRO/1000 Network Connection"),
128 	PVID(0x8086, E1000_DEV_ID_ICH8_82567V_3, "Intel(R) PRO/1000 Network Connection"),
129 	PVID(0x8086, E1000_DEV_ID_ICH9_IGP_M_AMT, "Intel(R) PRO/1000 Network Connection"),
130 	PVID(0x8086, E1000_DEV_ID_ICH9_IGP_AMT, "Intel(R) PRO/1000 Network Connection"),
131 	PVID(0x8086, E1000_DEV_ID_ICH9_IGP_C, "Intel(R) PRO/1000 Network Connection"),
132 	PVID(0x8086, E1000_DEV_ID_ICH9_IGP_M, "Intel(R) PRO/1000 Network Connection"),
133 	PVID(0x8086, E1000_DEV_ID_ICH9_IGP_M_V, "Intel(R) PRO/1000 Network Connection"),
134 	PVID(0x8086, E1000_DEV_ID_ICH9_IFE, "Intel(R) PRO/1000 Network Connection"),
135 	PVID(0x8086, E1000_DEV_ID_ICH9_IFE_GT, "Intel(R) PRO/1000 Network Connection"),
136 	PVID(0x8086, E1000_DEV_ID_ICH9_IFE_G, "Intel(R) PRO/1000 Network Connection"),
137 	PVID(0x8086, E1000_DEV_ID_ICH9_BM, "Intel(R) PRO/1000 Network Connection"),
138 	PVID(0x8086, E1000_DEV_ID_82574L, "Intel(R) PRO/1000 Network Connection"),
139 	PVID(0x8086, E1000_DEV_ID_82574LA, "Intel(R) PRO/1000 Network Connection"),
140 	PVID(0x8086, E1000_DEV_ID_ICH10_R_BM_LM, "Intel(R) PRO/1000 Network Connection"),
141 	PVID(0x8086, E1000_DEV_ID_ICH10_R_BM_LF, "Intel(R) PRO/1000 Network Connection"),
142 	PVID(0x8086, E1000_DEV_ID_ICH10_R_BM_V, "Intel(R) PRO/1000 Network Connection"),
143 	PVID(0x8086, E1000_DEV_ID_ICH10_D_BM_LM, "Intel(R) PRO/1000 Network Connection"),
144 	PVID(0x8086, E1000_DEV_ID_ICH10_D_BM_LF, "Intel(R) PRO/1000 Network Connection"),
145 	PVID(0x8086, E1000_DEV_ID_ICH10_D_BM_V, "Intel(R) PRO/1000 Network Connection"),
146 	PVID(0x8086, E1000_DEV_ID_PCH_M_HV_LM, "Intel(R) PRO/1000 Network Connection"),
147 	PVID(0x8086, E1000_DEV_ID_PCH_M_HV_LC, "Intel(R) PRO/1000 Network Connection"),
148 	PVID(0x8086, E1000_DEV_ID_PCH_D_HV_DM, "Intel(R) PRO/1000 Network Connection"),
149 	PVID(0x8086, E1000_DEV_ID_PCH_D_HV_DC, "Intel(R) PRO/1000 Network Connection"),
150 	PVID(0x8086, E1000_DEV_ID_PCH2_LV_LM, "Intel(R) PRO/1000 Network Connection"),
151 	PVID(0x8086, E1000_DEV_ID_PCH2_LV_V, "Intel(R) PRO/1000 Network Connection"),
152 	PVID(0x8086, E1000_DEV_ID_PCH_LPT_I217_LM, "Intel(R) PRO/1000 Network Connection"),
153 	PVID(0x8086, E1000_DEV_ID_PCH_LPT_I217_V, "Intel(R) PRO/1000 Network Connection"),
154 	PVID(0x8086, E1000_DEV_ID_PCH_LPTLP_I218_LM, "Intel(R) PRO/1000 Network Connection"),
155 	PVID(0x8086, E1000_DEV_ID_PCH_LPTLP_I218_V, "Intel(R) PRO/1000 Network Connection"),
156 	PVID(0x8086, E1000_DEV_ID_PCH_I218_LM2, "Intel(R) PRO/1000 Network Connection"),
157 	PVID(0x8086, E1000_DEV_ID_PCH_I218_V2, "Intel(R) PRO/1000 Network Connection"),
158 	PVID(0x8086, E1000_DEV_ID_PCH_I218_LM3, "Intel(R) PRO/1000 Network Connection"),
159 	PVID(0x8086, E1000_DEV_ID_PCH_I218_V3, "Intel(R) PRO/1000 Network Connection"),
160 	PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_LM, "Intel(R) PRO/1000 Network Connection"),
161 	PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_V, "Intel(R) PRO/1000 Network Connection"),
162 	PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_LM2, "Intel(R) PRO/1000 Network Connection"),
163 	PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_V2, "Intel(R) PRO/1000 Network Connection"),
164 	PVID(0x8086, E1000_DEV_ID_PCH_LBG_I219_LM3, "Intel(R) PRO/1000 Network Connection"),
165 	PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_LM4, "Intel(R) PRO/1000 Network Connection"),
166 	PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_V4, "Intel(R) PRO/1000 Network Connection"),
167 	PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_LM5, "Intel(R) PRO/1000 Network Connection"),
168 	PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_V5, "Intel(R) PRO/1000 Network Connection"),
169 	PVID(0x8086, E1000_DEV_ID_PCH_CNP_I219_LM6, "Intel(R) PRO/1000 Network Connection"),
170 	PVID(0x8086, E1000_DEV_ID_PCH_CNP_I219_V6, "Intel(R) PRO/1000 Network Connection"),
171 	PVID(0x8086, E1000_DEV_ID_PCH_CNP_I219_LM7, "Intel(R) PRO/1000 Network Connection"),
172 	PVID(0x8086, E1000_DEV_ID_PCH_CNP_I219_V7, "Intel(R) PRO/1000 Network Connection"),
173 	PVID(0x8086, E1000_DEV_ID_PCH_ICP_I219_LM8, "Intel(R) PRO/1000 Network Connection"),
174 	PVID(0x8086, E1000_DEV_ID_PCH_ICP_I219_V8, "Intel(R) PRO/1000 Network Connection"),
175 	PVID(0x8086, E1000_DEV_ID_PCH_ICP_I219_LM9, "Intel(R) PRO/1000 Network Connection"),
176 	PVID(0x8086, E1000_DEV_ID_PCH_ICP_I219_V9, "Intel(R) PRO/1000 Network Connection"),
177 	/* required last entry */
178 	PVID_END
179 };
180 
181 static pci_vendor_info_t igb_vendor_info_array[] =
182 {
183 	/* Intel(R) PRO/1000 Network Connection - igb */
184 	PVID(0x8086, E1000_DEV_ID_82575EB_COPPER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
185 	PVID(0x8086, E1000_DEV_ID_82575EB_FIBER_SERDES, "Intel(R) PRO/1000 PCI-Express Network Driver"),
186 	PVID(0x8086, E1000_DEV_ID_82575GB_QUAD_COPPER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
187 	PVID(0x8086, E1000_DEV_ID_82576, "Intel(R) PRO/1000 PCI-Express Network Driver"),
188 	PVID(0x8086, E1000_DEV_ID_82576_NS, "Intel(R) PRO/1000 PCI-Express Network Driver"),
189 	PVID(0x8086, E1000_DEV_ID_82576_NS_SERDES, "Intel(R) PRO/1000 PCI-Express Network Driver"),
190 	PVID(0x8086, E1000_DEV_ID_82576_FIBER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
191 	PVID(0x8086, E1000_DEV_ID_82576_SERDES, "Intel(R) PRO/1000 PCI-Express Network Driver"),
192 	PVID(0x8086, E1000_DEV_ID_82576_SERDES_QUAD, "Intel(R) PRO/1000 PCI-Express Network Driver"),
193 	PVID(0x8086, E1000_DEV_ID_82576_QUAD_COPPER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
194 	PVID(0x8086, E1000_DEV_ID_82576_QUAD_COPPER_ET2, "Intel(R) PRO/1000 PCI-Express Network Driver"),
195 	PVID(0x8086, E1000_DEV_ID_82576_VF, "Intel(R) PRO/1000 PCI-Express Network Driver"),
196 	PVID(0x8086, E1000_DEV_ID_82580_COPPER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
197 	PVID(0x8086, E1000_DEV_ID_82580_FIBER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
198 	PVID(0x8086, E1000_DEV_ID_82580_SERDES, "Intel(R) PRO/1000 PCI-Express Network Driver"),
199 	PVID(0x8086, E1000_DEV_ID_82580_SGMII, "Intel(R) PRO/1000 PCI-Express Network Driver"),
200 	PVID(0x8086, E1000_DEV_ID_82580_COPPER_DUAL, "Intel(R) PRO/1000 PCI-Express Network Driver"),
201 	PVID(0x8086, E1000_DEV_ID_82580_QUAD_FIBER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
202 	PVID(0x8086, E1000_DEV_ID_DH89XXCC_SERDES, "Intel(R) PRO/1000 PCI-Express Network Driver"),
203 	PVID(0x8086, E1000_DEV_ID_DH89XXCC_SGMII, "Intel(R) PRO/1000 PCI-Express Network Driver"),
204 	PVID(0x8086, E1000_DEV_ID_DH89XXCC_SFP, "Intel(R) PRO/1000 PCI-Express Network Driver"),
205 	PVID(0x8086, E1000_DEV_ID_DH89XXCC_BACKPLANE, "Intel(R) PRO/1000 PCI-Express Network Driver"),
206 	PVID(0x8086, E1000_DEV_ID_I350_COPPER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
207 	PVID(0x8086, E1000_DEV_ID_I350_FIBER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
208 	PVID(0x8086, E1000_DEV_ID_I350_SERDES, "Intel(R) PRO/1000 PCI-Express Network Driver"),
209 	PVID(0x8086, E1000_DEV_ID_I350_SGMII, "Intel(R) PRO/1000 PCI-Express Network Driver"),
210 	PVID(0x8086, E1000_DEV_ID_I350_VF, "Intel(R) PRO/1000 PCI-Express Network Driver"),
211 	PVID(0x8086, E1000_DEV_ID_I210_COPPER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
212 	PVID(0x8086, E1000_DEV_ID_I210_COPPER_IT, "Intel(R) PRO/1000 PCI-Express Network Driver"),
213 	PVID(0x8086, E1000_DEV_ID_I210_COPPER_OEM1, "Intel(R) PRO/1000 PCI-Express Network Driver"),
214 	PVID(0x8086, E1000_DEV_ID_I210_COPPER_FLASHLESS, "Intel(R) PRO/1000 PCI-Express Network Driver"),
215 	PVID(0x8086, E1000_DEV_ID_I210_SERDES_FLASHLESS, "Intel(R) PRO/1000 PCI-Express Network Driver"),
216 	PVID(0x8086, E1000_DEV_ID_I210_FIBER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
217 	PVID(0x8086, E1000_DEV_ID_I210_SERDES, "Intel(R) PRO/1000 PCI-Express Network Driver"),
218 	PVID(0x8086, E1000_DEV_ID_I210_SGMII, "Intel(R) PRO/1000 PCI-Express Network Driver"),
219 	PVID(0x8086, E1000_DEV_ID_I211_COPPER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
220 	PVID(0x8086, E1000_DEV_ID_I354_BACKPLANE_1GBPS, "Intel(R) PRO/1000 PCI-Express Network Driver"),
221 	PVID(0x8086, E1000_DEV_ID_I354_BACKPLANE_2_5GBPS, "Intel(R) PRO/1000 PCI-Express Network Driver"),
222 	PVID(0x8086, E1000_DEV_ID_I354_SGMII, "Intel(R) PRO/1000 PCI-Express Network Driver"),
223 	/* required last entry */
224 	PVID_END
225 };
226 
227 /*********************************************************************
228  *  Function prototypes
229  *********************************************************************/
230 static void	*em_register(device_t dev);
231 static void	*igb_register(device_t dev);
232 static int	em_if_attach_pre(if_ctx_t ctx);
233 static int	em_if_attach_post(if_ctx_t ctx);
234 static int	em_if_detach(if_ctx_t ctx);
235 static int	em_if_shutdown(if_ctx_t ctx);
236 static int	em_if_suspend(if_ctx_t ctx);
237 static int	em_if_resume(if_ctx_t ctx);
238 
239 static int	em_if_tx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int ntxqs, int ntxqsets);
240 static int	em_if_rx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int nrxqs, int nrxqsets);
241 static void	em_if_queues_free(if_ctx_t ctx);
242 
243 static uint64_t	em_if_get_counter(if_ctx_t, ift_counter);
244 static void	em_if_init(if_ctx_t ctx);
245 static void	em_if_stop(if_ctx_t ctx);
246 static void	em_if_media_status(if_ctx_t, struct ifmediareq *);
247 static int	em_if_media_change(if_ctx_t ctx);
248 static int	em_if_mtu_set(if_ctx_t ctx, uint32_t mtu);
249 static void	em_if_timer(if_ctx_t ctx, uint16_t qid);
250 static void	em_if_vlan_register(if_ctx_t ctx, u16 vtag);
251 static void	em_if_vlan_unregister(if_ctx_t ctx, u16 vtag);
252 static void	em_if_watchdog_reset(if_ctx_t ctx);
253 
254 static void	em_identify_hardware(if_ctx_t ctx);
255 static int	em_allocate_pci_resources(if_ctx_t ctx);
256 static void	em_free_pci_resources(if_ctx_t ctx);
257 static void	em_reset(if_ctx_t ctx);
258 static int	em_setup_interface(if_ctx_t ctx);
259 static int	em_setup_msix(if_ctx_t ctx);
260 
261 static void	em_initialize_transmit_unit(if_ctx_t ctx);
262 static void	em_initialize_receive_unit(if_ctx_t ctx);
263 
264 static void	em_if_enable_intr(if_ctx_t ctx);
265 static void	em_if_disable_intr(if_ctx_t ctx);
266 static int	em_if_rx_queue_intr_enable(if_ctx_t ctx, uint16_t rxqid);
267 static int	em_if_tx_queue_intr_enable(if_ctx_t ctx, uint16_t txqid);
268 static void	em_if_multi_set(if_ctx_t ctx);
269 static void	em_if_update_admin_status(if_ctx_t ctx);
270 static void	em_if_debug(if_ctx_t ctx);
271 static void	em_update_stats_counters(struct adapter *);
272 static void	em_add_hw_stats(struct adapter *adapter);
273 static int	em_if_set_promisc(if_ctx_t ctx, int flags);
274 static void	em_setup_vlan_hw_support(struct adapter *);
275 static int	em_sysctl_nvm_info(SYSCTL_HANDLER_ARGS);
276 static void	em_print_nvm_info(struct adapter *);
277 static int	em_sysctl_debug_info(SYSCTL_HANDLER_ARGS);
278 static int	em_get_rs(SYSCTL_HANDLER_ARGS);
279 static void	em_print_debug_info(struct adapter *);
280 static int 	em_is_valid_ether_addr(u8 *);
281 static int	em_sysctl_int_delay(SYSCTL_HANDLER_ARGS);
282 static void	em_add_int_delay_sysctl(struct adapter *, const char *,
283 		    const char *, struct em_int_delay_info *, int, int);
284 /* Management and WOL Support */
285 static void	em_init_manageability(struct adapter *);
286 static void	em_release_manageability(struct adapter *);
287 static void	em_get_hw_control(struct adapter *);
288 static void	em_release_hw_control(struct adapter *);
289 static void	em_get_wakeup(if_ctx_t ctx);
290 static void	em_enable_wakeup(if_ctx_t ctx);
291 static int	em_enable_phy_wakeup(struct adapter *);
292 static void	em_disable_aspm(struct adapter *);
293 
294 int		em_intr(void *arg);
295 static void	em_disable_promisc(if_ctx_t ctx);
296 
297 /* MSI-X handlers */
298 static int	em_if_msix_intr_assign(if_ctx_t, int);
299 static int	em_msix_link(void *);
300 static void	em_handle_link(void *context);
301 
302 static void	em_enable_vectors_82574(if_ctx_t);
303 
304 static int	em_set_flowcntl(SYSCTL_HANDLER_ARGS);
305 static int	em_sysctl_eee(SYSCTL_HANDLER_ARGS);
306 static void	em_if_led_func(if_ctx_t ctx, int onoff);
307 
308 static int	em_get_regs(SYSCTL_HANDLER_ARGS);
309 
310 static void	lem_smartspeed(struct adapter *adapter);
311 static void	igb_configure_queues(struct adapter *adapter);
312 
313 
314 /*********************************************************************
315  *  FreeBSD Device Interface Entry Points
316  *********************************************************************/
317 static device_method_t em_methods[] = {
318 	/* Device interface */
319 	DEVMETHOD(device_register, em_register),
320 	DEVMETHOD(device_probe, iflib_device_probe),
321 	DEVMETHOD(device_attach, iflib_device_attach),
322 	DEVMETHOD(device_detach, iflib_device_detach),
323 	DEVMETHOD(device_shutdown, iflib_device_shutdown),
324 	DEVMETHOD(device_suspend, iflib_device_suspend),
325 	DEVMETHOD(device_resume, iflib_device_resume),
326 	DEVMETHOD_END
327 };
328 
329 static device_method_t igb_methods[] = {
330 	/* Device interface */
331 	DEVMETHOD(device_register, igb_register),
332 	DEVMETHOD(device_probe, iflib_device_probe),
333 	DEVMETHOD(device_attach, iflib_device_attach),
334 	DEVMETHOD(device_detach, iflib_device_detach),
335 	DEVMETHOD(device_shutdown, iflib_device_shutdown),
336 	DEVMETHOD(device_suspend, iflib_device_suspend),
337 	DEVMETHOD(device_resume, iflib_device_resume),
338 	DEVMETHOD_END
339 };
340 
341 
342 static driver_t em_driver = {
343 	"em", em_methods, sizeof(struct adapter),
344 };
345 
346 static devclass_t em_devclass;
347 DRIVER_MODULE(em, pci, em_driver, em_devclass, 0, 0);
348 
349 MODULE_DEPEND(em, pci, 1, 1, 1);
350 MODULE_DEPEND(em, ether, 1, 1, 1);
351 MODULE_DEPEND(em, iflib, 1, 1, 1);
352 
353 IFLIB_PNP_INFO(pci, em, em_vendor_info_array);
354 
355 static driver_t igb_driver = {
356 	"igb", igb_methods, sizeof(struct adapter),
357 };
358 
359 static devclass_t igb_devclass;
360 DRIVER_MODULE(igb, pci, igb_driver, igb_devclass, 0, 0);
361 
362 MODULE_DEPEND(igb, pci, 1, 1, 1);
363 MODULE_DEPEND(igb, ether, 1, 1, 1);
364 MODULE_DEPEND(igb, iflib, 1, 1, 1);
365 
366 IFLIB_PNP_INFO(pci, igb, igb_vendor_info_array);
367 
368 static device_method_t em_if_methods[] = {
369 	DEVMETHOD(ifdi_attach_pre, em_if_attach_pre),
370 	DEVMETHOD(ifdi_attach_post, em_if_attach_post),
371 	DEVMETHOD(ifdi_detach, em_if_detach),
372 	DEVMETHOD(ifdi_shutdown, em_if_shutdown),
373 	DEVMETHOD(ifdi_suspend, em_if_suspend),
374 	DEVMETHOD(ifdi_resume, em_if_resume),
375 	DEVMETHOD(ifdi_init, em_if_init),
376 	DEVMETHOD(ifdi_stop, em_if_stop),
377 	DEVMETHOD(ifdi_msix_intr_assign, em_if_msix_intr_assign),
378 	DEVMETHOD(ifdi_intr_enable, em_if_enable_intr),
379 	DEVMETHOD(ifdi_intr_disable, em_if_disable_intr),
380 	DEVMETHOD(ifdi_tx_queues_alloc, em_if_tx_queues_alloc),
381 	DEVMETHOD(ifdi_rx_queues_alloc, em_if_rx_queues_alloc),
382 	DEVMETHOD(ifdi_queues_free, em_if_queues_free),
383 	DEVMETHOD(ifdi_update_admin_status, em_if_update_admin_status),
384 	DEVMETHOD(ifdi_multi_set, em_if_multi_set),
385 	DEVMETHOD(ifdi_media_status, em_if_media_status),
386 	DEVMETHOD(ifdi_media_change, em_if_media_change),
387 	DEVMETHOD(ifdi_mtu_set, em_if_mtu_set),
388 	DEVMETHOD(ifdi_promisc_set, em_if_set_promisc),
389 	DEVMETHOD(ifdi_timer, em_if_timer),
390 	DEVMETHOD(ifdi_watchdog_reset, em_if_watchdog_reset),
391 	DEVMETHOD(ifdi_vlan_register, em_if_vlan_register),
392 	DEVMETHOD(ifdi_vlan_unregister, em_if_vlan_unregister),
393 	DEVMETHOD(ifdi_get_counter, em_if_get_counter),
394 	DEVMETHOD(ifdi_led_func, em_if_led_func),
395 	DEVMETHOD(ifdi_rx_queue_intr_enable, em_if_rx_queue_intr_enable),
396 	DEVMETHOD(ifdi_tx_queue_intr_enable, em_if_tx_queue_intr_enable),
397 	DEVMETHOD(ifdi_debug, em_if_debug),
398 	DEVMETHOD_END
399 };
400 
401 /*
402  * note that if (adapter->msix_mem) is replaced by:
403  * if (adapter->intr_type == IFLIB_INTR_MSIX)
404  */
405 static driver_t em_if_driver = {
406 	"em_if", em_if_methods, sizeof(struct adapter)
407 };
408 
409 /*********************************************************************
410  *  Tunable default values.
411  *********************************************************************/
412 
413 #define EM_TICKS_TO_USECS(ticks)	((1024 * (ticks) + 500) / 1000)
414 #define EM_USECS_TO_TICKS(usecs)	((1000 * (usecs) + 512) / 1024)
415 
416 #define MAX_INTS_PER_SEC	8000
417 #define DEFAULT_ITR		(1000000000/(MAX_INTS_PER_SEC * 256))
418 
419 /* Allow common code without TSO */
420 #ifndef CSUM_TSO
421 #define CSUM_TSO	0
422 #endif
423 
424 static SYSCTL_NODE(_hw, OID_AUTO, em, CTLFLAG_RD, 0, "EM driver parameters");
425 
426 static int em_disable_crc_stripping = 0;
427 SYSCTL_INT(_hw_em, OID_AUTO, disable_crc_stripping, CTLFLAG_RDTUN,
428     &em_disable_crc_stripping, 0, "Disable CRC Stripping");
429 
430 static int em_tx_int_delay_dflt = EM_TICKS_TO_USECS(EM_TIDV);
431 static int em_rx_int_delay_dflt = EM_TICKS_TO_USECS(EM_RDTR);
432 SYSCTL_INT(_hw_em, OID_AUTO, tx_int_delay, CTLFLAG_RDTUN, &em_tx_int_delay_dflt,
433     0, "Default transmit interrupt delay in usecs");
434 SYSCTL_INT(_hw_em, OID_AUTO, rx_int_delay, CTLFLAG_RDTUN, &em_rx_int_delay_dflt,
435     0, "Default receive interrupt delay in usecs");
436 
437 static int em_tx_abs_int_delay_dflt = EM_TICKS_TO_USECS(EM_TADV);
438 static int em_rx_abs_int_delay_dflt = EM_TICKS_TO_USECS(EM_RADV);
439 SYSCTL_INT(_hw_em, OID_AUTO, tx_abs_int_delay, CTLFLAG_RDTUN,
440     &em_tx_abs_int_delay_dflt, 0,
441     "Default transmit interrupt delay limit in usecs");
442 SYSCTL_INT(_hw_em, OID_AUTO, rx_abs_int_delay, CTLFLAG_RDTUN,
443     &em_rx_abs_int_delay_dflt, 0,
444     "Default receive interrupt delay limit in usecs");
445 
446 static int em_smart_pwr_down = FALSE;
447 SYSCTL_INT(_hw_em, OID_AUTO, smart_pwr_down, CTLFLAG_RDTUN, &em_smart_pwr_down,
448     0, "Set to true to leave smart power down enabled on newer adapters");
449 
450 /* Controls whether promiscuous also shows bad packets */
451 static int em_debug_sbp = TRUE;
452 SYSCTL_INT(_hw_em, OID_AUTO, sbp, CTLFLAG_RDTUN, &em_debug_sbp, 0,
453     "Show bad packets in promiscuous mode");
454 
455 /* How many packets rxeof tries to clean at a time */
456 static int em_rx_process_limit = 100;
457 SYSCTL_INT(_hw_em, OID_AUTO, rx_process_limit, CTLFLAG_RDTUN,
458     &em_rx_process_limit, 0,
459     "Maximum number of received packets to process "
460     "at a time, -1 means unlimited");
461 
462 /* Energy efficient ethernet - default to OFF */
463 static int eee_setting = 1;
464 SYSCTL_INT(_hw_em, OID_AUTO, eee_setting, CTLFLAG_RDTUN, &eee_setting, 0,
465     "Enable Energy Efficient Ethernet");
466 
467 /*
468 ** Tuneable Interrupt rate
469 */
470 static int em_max_interrupt_rate = 8000;
471 SYSCTL_INT(_hw_em, OID_AUTO, max_interrupt_rate, CTLFLAG_RDTUN,
472     &em_max_interrupt_rate, 0, "Maximum interrupts per second");
473 
474 
475 
476 /* Global used in WOL setup with multiport cards */
477 static int global_quad_port_a = 0;
478 
479 extern struct if_txrx igb_txrx;
480 extern struct if_txrx em_txrx;
481 extern struct if_txrx lem_txrx;
482 
483 static struct if_shared_ctx em_sctx_init = {
484 	.isc_magic = IFLIB_MAGIC,
485 	.isc_q_align = PAGE_SIZE,
486 	.isc_tx_maxsize = EM_TSO_SIZE + sizeof(struct ether_vlan_header),
487 	.isc_tx_maxsegsize = PAGE_SIZE,
488 	.isc_tso_maxsize = EM_TSO_SIZE + sizeof(struct ether_vlan_header),
489 	.isc_tso_maxsegsize = EM_TSO_SEG_SIZE,
490 	.isc_rx_maxsize = MJUM9BYTES,
491 	.isc_rx_nsegments = 1,
492 	.isc_rx_maxsegsize = MJUM9BYTES,
493 	.isc_nfl = 1,
494 	.isc_nrxqs = 1,
495 	.isc_ntxqs = 1,
496 	.isc_admin_intrcnt = 1,
497 	.isc_vendor_info = em_vendor_info_array,
498 	.isc_driver_version = em_driver_version,
499 	.isc_driver = &em_if_driver,
500 	.isc_flags = IFLIB_NEED_SCRATCH | IFLIB_TSO_INIT_IP | IFLIB_NEED_ZERO_CSUM,
501 
502 	.isc_nrxd_min = {EM_MIN_RXD},
503 	.isc_ntxd_min = {EM_MIN_TXD},
504 	.isc_nrxd_max = {EM_MAX_RXD},
505 	.isc_ntxd_max = {EM_MAX_TXD},
506 	.isc_nrxd_default = {EM_DEFAULT_RXD},
507 	.isc_ntxd_default = {EM_DEFAULT_TXD},
508 };
509 
510 if_shared_ctx_t em_sctx = &em_sctx_init;
511 
512 static struct if_shared_ctx igb_sctx_init = {
513 	.isc_magic = IFLIB_MAGIC,
514 	.isc_q_align = PAGE_SIZE,
515 	.isc_tx_maxsize = EM_TSO_SIZE + sizeof(struct ether_vlan_header),
516 	.isc_tx_maxsegsize = PAGE_SIZE,
517 	.isc_tso_maxsize = EM_TSO_SIZE + sizeof(struct ether_vlan_header),
518 	.isc_tso_maxsegsize = EM_TSO_SEG_SIZE,
519 	.isc_rx_maxsize = MJUM9BYTES,
520 	.isc_rx_nsegments = 1,
521 	.isc_rx_maxsegsize = MJUM9BYTES,
522 	.isc_nfl = 1,
523 	.isc_nrxqs = 1,
524 	.isc_ntxqs = 1,
525 	.isc_admin_intrcnt = 1,
526 	.isc_vendor_info = igb_vendor_info_array,
527 	.isc_driver_version = em_driver_version,
528 	.isc_driver = &em_if_driver,
529 	.isc_flags = IFLIB_NEED_SCRATCH | IFLIB_TSO_INIT_IP | IFLIB_NEED_ZERO_CSUM,
530 
531 	.isc_nrxd_min = {EM_MIN_RXD},
532 	.isc_ntxd_min = {EM_MIN_TXD},
533 	.isc_nrxd_max = {IGB_MAX_RXD},
534 	.isc_ntxd_max = {IGB_MAX_TXD},
535 	.isc_nrxd_default = {EM_DEFAULT_RXD},
536 	.isc_ntxd_default = {EM_DEFAULT_TXD},
537 };
538 
539 if_shared_ctx_t igb_sctx = &igb_sctx_init;
540 
541 /*****************************************************************
542  *
543  * Dump Registers
544  *
545  ****************************************************************/
546 #define IGB_REGS_LEN 739
547 
548 static int em_get_regs(SYSCTL_HANDLER_ARGS)
549 {
550 	struct adapter *adapter = (struct adapter *)arg1;
551 	struct e1000_hw *hw = &adapter->hw;
552 	struct sbuf *sb;
553 	u32 *regs_buff;
554 	int rc;
555 
556 	regs_buff = malloc(sizeof(u32) * IGB_REGS_LEN, M_DEVBUF, M_WAITOK);
557 	memset(regs_buff, 0, IGB_REGS_LEN * sizeof(u32));
558 
559 	rc = sysctl_wire_old_buffer(req, 0);
560 	MPASS(rc == 0);
561 	if (rc != 0) {
562 		free(regs_buff, M_DEVBUF);
563 		return (rc);
564 	}
565 
566 	sb = sbuf_new_for_sysctl(NULL, NULL, 32*400, req);
567 	MPASS(sb != NULL);
568 	if (sb == NULL) {
569 		free(regs_buff, M_DEVBUF);
570 		return (ENOMEM);
571 	}
572 
573 	/* General Registers */
574 	regs_buff[0] = E1000_READ_REG(hw, E1000_CTRL);
575 	regs_buff[1] = E1000_READ_REG(hw, E1000_STATUS);
576 	regs_buff[2] = E1000_READ_REG(hw, E1000_CTRL_EXT);
577 	regs_buff[3] = E1000_READ_REG(hw, E1000_ICR);
578 	regs_buff[4] = E1000_READ_REG(hw, E1000_RCTL);
579 	regs_buff[5] = E1000_READ_REG(hw, E1000_RDLEN(0));
580 	regs_buff[6] = E1000_READ_REG(hw, E1000_RDH(0));
581 	regs_buff[7] = E1000_READ_REG(hw, E1000_RDT(0));
582 	regs_buff[8] = E1000_READ_REG(hw, E1000_RXDCTL(0));
583 	regs_buff[9] = E1000_READ_REG(hw, E1000_RDBAL(0));
584 	regs_buff[10] = E1000_READ_REG(hw, E1000_RDBAH(0));
585 	regs_buff[11] = E1000_READ_REG(hw, E1000_TCTL);
586 	regs_buff[12] = E1000_READ_REG(hw, E1000_TDBAL(0));
587 	regs_buff[13] = E1000_READ_REG(hw, E1000_TDBAH(0));
588 	regs_buff[14] = E1000_READ_REG(hw, E1000_TDLEN(0));
589 	regs_buff[15] = E1000_READ_REG(hw, E1000_TDH(0));
590 	regs_buff[16] = E1000_READ_REG(hw, E1000_TDT(0));
591 	regs_buff[17] = E1000_READ_REG(hw, E1000_TXDCTL(0));
592 	regs_buff[18] = E1000_READ_REG(hw, E1000_TDFH);
593 	regs_buff[19] = E1000_READ_REG(hw, E1000_TDFT);
594 	regs_buff[20] = E1000_READ_REG(hw, E1000_TDFHS);
595 	regs_buff[21] = E1000_READ_REG(hw, E1000_TDFPC);
596 
597 	sbuf_printf(sb, "General Registers\n");
598 	sbuf_printf(sb, "\tCTRL\t %08x\n", regs_buff[0]);
599 	sbuf_printf(sb, "\tSTATUS\t %08x\n", regs_buff[1]);
600 	sbuf_printf(sb, "\tCTRL_EXIT\t %08x\n\n", regs_buff[2]);
601 
602 	sbuf_printf(sb, "Interrupt Registers\n");
603 	sbuf_printf(sb, "\tICR\t %08x\n\n", regs_buff[3]);
604 
605 	sbuf_printf(sb, "RX Registers\n");
606 	sbuf_printf(sb, "\tRCTL\t %08x\n", regs_buff[4]);
607 	sbuf_printf(sb, "\tRDLEN\t %08x\n", regs_buff[5]);
608 	sbuf_printf(sb, "\tRDH\t %08x\n", regs_buff[6]);
609 	sbuf_printf(sb, "\tRDT\t %08x\n", regs_buff[7]);
610 	sbuf_printf(sb, "\tRXDCTL\t %08x\n", regs_buff[8]);
611 	sbuf_printf(sb, "\tRDBAL\t %08x\n", regs_buff[9]);
612 	sbuf_printf(sb, "\tRDBAH\t %08x\n\n", regs_buff[10]);
613 
614 	sbuf_printf(sb, "TX Registers\n");
615 	sbuf_printf(sb, "\tTCTL\t %08x\n", regs_buff[11]);
616 	sbuf_printf(sb, "\tTDBAL\t %08x\n", regs_buff[12]);
617 	sbuf_printf(sb, "\tTDBAH\t %08x\n", regs_buff[13]);
618 	sbuf_printf(sb, "\tTDLEN\t %08x\n", regs_buff[14]);
619 	sbuf_printf(sb, "\tTDH\t %08x\n", regs_buff[15]);
620 	sbuf_printf(sb, "\tTDT\t %08x\n", regs_buff[16]);
621 	sbuf_printf(sb, "\tTXDCTL\t %08x\n", regs_buff[17]);
622 	sbuf_printf(sb, "\tTDFH\t %08x\n", regs_buff[18]);
623 	sbuf_printf(sb, "\tTDFT\t %08x\n", regs_buff[19]);
624 	sbuf_printf(sb, "\tTDFHS\t %08x\n", regs_buff[20]);
625 	sbuf_printf(sb, "\tTDFPC\t %08x\n\n", regs_buff[21]);
626 
627 	free(regs_buff, M_DEVBUF);
628 
629 #ifdef DUMP_DESCS
630 	{
631 		if_softc_ctx_t scctx = adapter->shared;
632 		struct rx_ring *rxr = &rx_que->rxr;
633 		struct tx_ring *txr = &tx_que->txr;
634 		int ntxd = scctx->isc_ntxd[0];
635 		int nrxd = scctx->isc_nrxd[0];
636 		int j;
637 
638 	for (j = 0; j < nrxd; j++) {
639 		u32 staterr = le32toh(rxr->rx_base[j].wb.upper.status_error);
640 		u32 length =  le32toh(rxr->rx_base[j].wb.upper.length);
641 		sbuf_printf(sb, "\tReceive Descriptor Address %d: %08" PRIx64 "  Error:%d  Length:%d\n", j, rxr->rx_base[j].read.buffer_addr, staterr, length);
642 	}
643 
644 	for (j = 0; j < min(ntxd, 256); j++) {
645 		unsigned int *ptr = (unsigned int *)&txr->tx_base[j];
646 
647 		sbuf_printf(sb, "\tTXD[%03d] [0]: %08x [1]: %08x [2]: %08x [3]: %08x  eop: %d DD=%d\n",
648 			    j, ptr[0], ptr[1], ptr[2], ptr[3], buf->eop,
649 			    buf->eop != -1 ? txr->tx_base[buf->eop].upper.fields.status & E1000_TXD_STAT_DD : 0);
650 
651 	}
652 	}
653 #endif
654 
655 	rc = sbuf_finish(sb);
656 	sbuf_delete(sb);
657 	return(rc);
658 }
659 
660 static void *
661 em_register(device_t dev)
662 {
663 	return (em_sctx);
664 }
665 
666 static void *
667 igb_register(device_t dev)
668 {
669 	return (igb_sctx);
670 }
671 
672 static int
673 em_set_num_queues(if_ctx_t ctx)
674 {
675 	struct adapter *adapter = iflib_get_softc(ctx);
676 	int maxqueues;
677 
678 	/* Sanity check based on HW */
679 	switch (adapter->hw.mac.type) {
680 	case e1000_82576:
681 	case e1000_82580:
682 	case e1000_i350:
683 	case e1000_i354:
684 		maxqueues = 8;
685 		break;
686 	case e1000_i210:
687 	case e1000_82575:
688 		maxqueues = 4;
689 		break;
690 	case e1000_i211:
691 	case e1000_82574:
692 		maxqueues = 2;
693 		break;
694 	default:
695 		maxqueues = 1;
696 		break;
697 	}
698 
699 	return (maxqueues);
700 }
701 
702 #define	LEM_CAPS							\
703     IFCAP_HWCSUM | IFCAP_VLAN_MTU | IFCAP_VLAN_HWTAGGING |		\
704     IFCAP_VLAN_HWCSUM | IFCAP_WOL | IFCAP_VLAN_HWFILTER
705 
706 #define	EM_CAPS								\
707     IFCAP_HWCSUM | IFCAP_VLAN_MTU | IFCAP_VLAN_HWTAGGING |		\
708     IFCAP_VLAN_HWCSUM | IFCAP_WOL | IFCAP_VLAN_HWFILTER | IFCAP_TSO4 |	\
709     IFCAP_LRO | IFCAP_VLAN_HWTSO
710 
711 #define	IGB_CAPS							\
712     IFCAP_HWCSUM | IFCAP_VLAN_MTU | IFCAP_VLAN_HWTAGGING |		\
713     IFCAP_VLAN_HWCSUM | IFCAP_WOL | IFCAP_VLAN_HWFILTER | IFCAP_TSO4 |	\
714     IFCAP_LRO | IFCAP_VLAN_HWTSO | IFCAP_JUMBO_MTU | IFCAP_HWCSUM_IPV6 |\
715     IFCAP_TSO6
716 
717 /*********************************************************************
718  *  Device initialization routine
719  *
720  *  The attach entry point is called when the driver is being loaded.
721  *  This routine identifies the type of hardware, allocates all resources
722  *  and initializes the hardware.
723  *
724  *  return 0 on success, positive on failure
725  *********************************************************************/
726 static int
727 em_if_attach_pre(if_ctx_t ctx)
728 {
729 	struct adapter *adapter;
730 	if_softc_ctx_t scctx;
731 	device_t dev;
732 	struct e1000_hw *hw;
733 	int error = 0;
734 
735 	INIT_DEBUGOUT("em_if_attach_pre: begin");
736 	dev = iflib_get_dev(ctx);
737 	adapter = iflib_get_softc(ctx);
738 
739 	adapter->ctx = adapter->osdep.ctx = ctx;
740 	adapter->dev = adapter->osdep.dev = dev;
741 	scctx = adapter->shared = iflib_get_softc_ctx(ctx);
742 	adapter->media = iflib_get_media(ctx);
743 	hw = &adapter->hw;
744 
745 	adapter->tx_process_limit = scctx->isc_ntxd[0];
746 
747 	/* SYSCTL stuff */
748 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
749 	    SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
750 	    OID_AUTO, "nvm", CTLTYPE_INT|CTLFLAG_RW, adapter, 0,
751 	    em_sysctl_nvm_info, "I", "NVM Information");
752 
753 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
754 	    SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
755 	    OID_AUTO, "debug", CTLTYPE_INT|CTLFLAG_RW, adapter, 0,
756 	    em_sysctl_debug_info, "I", "Debug Information");
757 
758 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
759 	    SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
760 	    OID_AUTO, "fc", CTLTYPE_INT|CTLFLAG_RW, adapter, 0,
761 	    em_set_flowcntl, "I", "Flow Control");
762 
763 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
764 	    SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
765 	    OID_AUTO, "reg_dump", CTLTYPE_STRING | CTLFLAG_RD, adapter, 0,
766 	    em_get_regs, "A", "Dump Registers");
767 
768 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
769 	    SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
770 	    OID_AUTO, "rs_dump", CTLTYPE_INT | CTLFLAG_RW, adapter, 0,
771 	    em_get_rs, "I", "Dump RS indexes");
772 
773 	/* Determine hardware and mac info */
774 	em_identify_hardware(ctx);
775 
776 	scctx->isc_tx_nsegments = EM_MAX_SCATTER;
777 	scctx->isc_nrxqsets_max = scctx->isc_ntxqsets_max = em_set_num_queues(ctx);
778 	if (bootverbose)
779 		device_printf(dev, "attach_pre capping queues at %d\n",
780 		    scctx->isc_ntxqsets_max);
781 
782 	if (adapter->hw.mac.type >= igb_mac_min) {
783 		scctx->isc_txqsizes[0] = roundup2(scctx->isc_ntxd[0] * sizeof(union e1000_adv_tx_desc), EM_DBA_ALIGN);
784 		scctx->isc_rxqsizes[0] = roundup2(scctx->isc_nrxd[0] * sizeof(union e1000_adv_rx_desc), EM_DBA_ALIGN);
785 		scctx->isc_txd_size[0] = sizeof(union e1000_adv_tx_desc);
786 		scctx->isc_rxd_size[0] = sizeof(union e1000_adv_rx_desc);
787 		scctx->isc_txrx = &igb_txrx;
788 		scctx->isc_tx_tso_segments_max = EM_MAX_SCATTER;
789 		scctx->isc_tx_tso_size_max = EM_TSO_SIZE;
790 		scctx->isc_tx_tso_segsize_max = EM_TSO_SEG_SIZE;
791 		scctx->isc_capabilities = scctx->isc_capenable = IGB_CAPS;
792 		scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP | CSUM_TSO |
793 		     CSUM_IP6_TCP | CSUM_IP6_UDP;
794 		if (adapter->hw.mac.type != e1000_82575)
795 			scctx->isc_tx_csum_flags |= CSUM_SCTP | CSUM_IP6_SCTP;
796 		/*
797 		** Some new devices, as with ixgbe, now may
798 		** use a different BAR, so we need to keep
799 		** track of which is used.
800 		*/
801 		scctx->isc_msix_bar = PCIR_BAR(EM_MSIX_BAR);
802 		if (pci_read_config(dev, scctx->isc_msix_bar, 4) == 0)
803 			scctx->isc_msix_bar += 4;
804 	} else if (adapter->hw.mac.type >= em_mac_min) {
805 		scctx->isc_txqsizes[0] = roundup2(scctx->isc_ntxd[0]* sizeof(struct e1000_tx_desc), EM_DBA_ALIGN);
806 		scctx->isc_rxqsizes[0] = roundup2(scctx->isc_nrxd[0] * sizeof(union e1000_rx_desc_extended), EM_DBA_ALIGN);
807 		scctx->isc_txd_size[0] = sizeof(struct e1000_tx_desc);
808 		scctx->isc_rxd_size[0] = sizeof(union e1000_rx_desc_extended);
809 		scctx->isc_txrx = &em_txrx;
810 		scctx->isc_tx_tso_segments_max = EM_MAX_SCATTER;
811 		scctx->isc_tx_tso_size_max = EM_TSO_SIZE;
812 		scctx->isc_tx_tso_segsize_max = EM_TSO_SEG_SIZE;
813 		scctx->isc_capabilities = scctx->isc_capenable = EM_CAPS;
814 		/*
815 		 * For EM-class devices, don't enable IFCAP_{TSO4,VLAN_HWTSO}
816 		 * by default as we don't have workarounds for all associated
817 		 * silicon errata.  E. g., with several MACs such as 82573E,
818 		 * TSO only works at Gigabit speed and otherwise can cause the
819 		 * hardware to hang (which also would be next to impossible to
820 		 * work around given that already queued TSO-using descriptors
821 		 * would need to be flushed and vlan(4) reconfigured at runtime
822 		 * in case of a link speed change).  Moreover, MACs like 82579
823 		 * still can hang at Gigabit even with all publicly documented
824 		 * TSO workarounds implemented.  Generally, the penality of
825 		 * these workarounds is rather high and may involve copying
826 		 * mbuf data around so advantages of TSO lapse.  Still, TSO may
827 		 * work for a few MACs of this class - at least when sticking
828 		 * with Gigabit - in which case users may enable TSO manually.
829 		 */
830 		scctx->isc_capenable &= ~(IFCAP_TSO4 | IFCAP_VLAN_HWTSO);
831 		scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP | CSUM_IP_TSO;
832 		/*
833 		 * We support MSI-X with 82574 only, but indicate to iflib(4)
834 		 * that it shall give MSI at least a try with other devices.
835 		 */
836 		if (adapter->hw.mac.type == e1000_82574) {
837 			scctx->isc_msix_bar = PCIR_BAR(EM_MSIX_BAR);
838 		} else {
839 			scctx->isc_msix_bar = -1;
840 			scctx->isc_disable_msix = 1;
841 		}
842 	} else {
843 		scctx->isc_txqsizes[0] = roundup2((scctx->isc_ntxd[0] + 1) * sizeof(struct e1000_tx_desc), EM_DBA_ALIGN);
844 		scctx->isc_rxqsizes[0] = roundup2((scctx->isc_nrxd[0] + 1) * sizeof(struct e1000_rx_desc), EM_DBA_ALIGN);
845 		scctx->isc_txd_size[0] = sizeof(struct e1000_tx_desc);
846 		scctx->isc_rxd_size[0] = sizeof(struct e1000_rx_desc);
847 		scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP;
848 		scctx->isc_txrx = &lem_txrx;
849 		scctx->isc_capabilities = scctx->isc_capenable = LEM_CAPS;
850 		if (adapter->hw.mac.type < e1000_82543)
851 			scctx->isc_capenable &= ~(IFCAP_HWCSUM|IFCAP_VLAN_HWCSUM);
852 		/* INTx only */
853 		scctx->isc_msix_bar = 0;
854 	}
855 
856 	/* Setup PCI resources */
857 	if (em_allocate_pci_resources(ctx)) {
858 		device_printf(dev, "Allocation of PCI resources failed\n");
859 		error = ENXIO;
860 		goto err_pci;
861 	}
862 
863 	/*
864 	** For ICH8 and family we need to
865 	** map the flash memory, and this
866 	** must happen after the MAC is
867 	** identified
868 	*/
869 	if ((hw->mac.type == e1000_ich8lan) ||
870 	    (hw->mac.type == e1000_ich9lan) ||
871 	    (hw->mac.type == e1000_ich10lan) ||
872 	    (hw->mac.type == e1000_pchlan) ||
873 	    (hw->mac.type == e1000_pch2lan) ||
874 	    (hw->mac.type == e1000_pch_lpt)) {
875 		int rid = EM_BAR_TYPE_FLASH;
876 		adapter->flash = bus_alloc_resource_any(dev,
877 		    SYS_RES_MEMORY, &rid, RF_ACTIVE);
878 		if (adapter->flash == NULL) {
879 			device_printf(dev, "Mapping of Flash failed\n");
880 			error = ENXIO;
881 			goto err_pci;
882 		}
883 		/* This is used in the shared code */
884 		hw->flash_address = (u8 *)adapter->flash;
885 		adapter->osdep.flash_bus_space_tag =
886 		    rman_get_bustag(adapter->flash);
887 		adapter->osdep.flash_bus_space_handle =
888 		    rman_get_bushandle(adapter->flash);
889 	}
890 	/*
891 	** In the new SPT device flash is not  a
892 	** separate BAR, rather it is also in BAR0,
893 	** so use the same tag and an offset handle for the
894 	** FLASH read/write macros in the shared code.
895 	*/
896 	else if (hw->mac.type >= e1000_pch_spt) {
897 		adapter->osdep.flash_bus_space_tag =
898 		    adapter->osdep.mem_bus_space_tag;
899 		adapter->osdep.flash_bus_space_handle =
900 		    adapter->osdep.mem_bus_space_handle
901 		    + E1000_FLASH_BASE_ADDR;
902 	}
903 
904 	/* Do Shared Code initialization */
905 	error = e1000_setup_init_funcs(hw, TRUE);
906 	if (error) {
907 		device_printf(dev, "Setup of Shared code failed, error %d\n",
908 		    error);
909 		error = ENXIO;
910 		goto err_pci;
911 	}
912 
913 	em_setup_msix(ctx);
914 	e1000_get_bus_info(hw);
915 
916 	/* Set up some sysctls for the tunable interrupt delays */
917 	em_add_int_delay_sysctl(adapter, "rx_int_delay",
918 	    "receive interrupt delay in usecs", &adapter->rx_int_delay,
919 	    E1000_REGISTER(hw, E1000_RDTR), em_rx_int_delay_dflt);
920 	em_add_int_delay_sysctl(adapter, "tx_int_delay",
921 	    "transmit interrupt delay in usecs", &adapter->tx_int_delay,
922 	    E1000_REGISTER(hw, E1000_TIDV), em_tx_int_delay_dflt);
923 	em_add_int_delay_sysctl(adapter, "rx_abs_int_delay",
924 	    "receive interrupt delay limit in usecs",
925 	    &adapter->rx_abs_int_delay,
926 	    E1000_REGISTER(hw, E1000_RADV),
927 	    em_rx_abs_int_delay_dflt);
928 	em_add_int_delay_sysctl(adapter, "tx_abs_int_delay",
929 	    "transmit interrupt delay limit in usecs",
930 	    &adapter->tx_abs_int_delay,
931 	    E1000_REGISTER(hw, E1000_TADV),
932 	    em_tx_abs_int_delay_dflt);
933 	em_add_int_delay_sysctl(adapter, "itr",
934 	    "interrupt delay limit in usecs/4",
935 	    &adapter->tx_itr,
936 	    E1000_REGISTER(hw, E1000_ITR),
937 	    DEFAULT_ITR);
938 
939 	hw->mac.autoneg = DO_AUTO_NEG;
940 	hw->phy.autoneg_wait_to_complete = FALSE;
941 	hw->phy.autoneg_advertised = AUTONEG_ADV_DEFAULT;
942 
943 	if (adapter->hw.mac.type < em_mac_min) {
944 		e1000_init_script_state_82541(&adapter->hw, TRUE);
945 		e1000_set_tbi_compatibility_82543(&adapter->hw, TRUE);
946 	}
947 	/* Copper options */
948 	if (hw->phy.media_type == e1000_media_type_copper) {
949 		hw->phy.mdix = AUTO_ALL_MODES;
950 		hw->phy.disable_polarity_correction = FALSE;
951 		hw->phy.ms_type = EM_MASTER_SLAVE;
952 	}
953 
954 	/*
955 	 * Set the frame limits assuming
956 	 * standard ethernet sized frames.
957 	 */
958 	scctx->isc_max_frame_size = adapter->hw.mac.max_frame_size =
959 	    ETHERMTU + ETHER_HDR_LEN + ETHERNET_FCS_SIZE;
960 
961 	/*
962 	 * This controls when hardware reports transmit completion
963 	 * status.
964 	 */
965 	hw->mac.report_tx_early = 1;
966 
967 	/* Allocate multicast array memory. */
968 	adapter->mta = malloc(sizeof(u8) * ETH_ADDR_LEN *
969 	    MAX_NUM_MULTICAST_ADDRESSES, M_DEVBUF, M_NOWAIT);
970 	if (adapter->mta == NULL) {
971 		device_printf(dev, "Can not allocate multicast setup array\n");
972 		error = ENOMEM;
973 		goto err_late;
974 	}
975 
976 	/* Check SOL/IDER usage */
977 	if (e1000_check_reset_block(hw))
978 		device_printf(dev, "PHY reset is blocked"
979 			      " due to SOL/IDER session.\n");
980 
981 	/* Sysctl for setting Energy Efficient Ethernet */
982 	hw->dev_spec.ich8lan.eee_disable = eee_setting;
983 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
984 	    SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
985 	    OID_AUTO, "eee_control", CTLTYPE_INT|CTLFLAG_RW,
986 	    adapter, 0, em_sysctl_eee, "I",
987 	    "Disable Energy Efficient Ethernet");
988 
989 	/*
990 	** Start from a known state, this is
991 	** important in reading the nvm and
992 	** mac from that.
993 	*/
994 	e1000_reset_hw(hw);
995 
996 	/* Make sure we have a good EEPROM before we read from it */
997 	if (e1000_validate_nvm_checksum(hw) < 0) {
998 		/*
999 		** Some PCI-E parts fail the first check due to
1000 		** the link being in sleep state, call it again,
1001 		** if it fails a second time its a real issue.
1002 		*/
1003 		if (e1000_validate_nvm_checksum(hw) < 0) {
1004 			device_printf(dev,
1005 			    "The EEPROM Checksum Is Not Valid\n");
1006 			error = EIO;
1007 			goto err_late;
1008 		}
1009 	}
1010 
1011 	/* Copy the permanent MAC address out of the EEPROM */
1012 	if (e1000_read_mac_addr(hw) < 0) {
1013 		device_printf(dev, "EEPROM read error while reading MAC"
1014 			      " address\n");
1015 		error = EIO;
1016 		goto err_late;
1017 	}
1018 
1019 	if (!em_is_valid_ether_addr(hw->mac.addr)) {
1020 		device_printf(dev, "Invalid MAC address\n");
1021 		error = EIO;
1022 		goto err_late;
1023 	}
1024 
1025 	/* Disable ULP support */
1026 	e1000_disable_ulp_lpt_lp(hw, TRUE);
1027 
1028 	/*
1029 	 * Get Wake-on-Lan and Management info for later use
1030 	 */
1031 	em_get_wakeup(ctx);
1032 
1033 	/* Enable only WOL MAGIC by default */
1034 	scctx->isc_capenable &= ~IFCAP_WOL;
1035 	if (adapter->wol != 0)
1036 		scctx->isc_capenable |= IFCAP_WOL_MAGIC;
1037 
1038 	iflib_set_mac(ctx, hw->mac.addr);
1039 
1040 	return (0);
1041 
1042 err_late:
1043 	em_release_hw_control(adapter);
1044 err_pci:
1045 	em_free_pci_resources(ctx);
1046 	free(adapter->mta, M_DEVBUF);
1047 
1048 	return (error);
1049 }
1050 
1051 static int
1052 em_if_attach_post(if_ctx_t ctx)
1053 {
1054 	struct adapter *adapter = iflib_get_softc(ctx);
1055 	struct e1000_hw *hw = &adapter->hw;
1056 	int error = 0;
1057 
1058 	/* Setup OS specific network interface */
1059 	error = em_setup_interface(ctx);
1060 	if (error != 0) {
1061 		goto err_late;
1062 	}
1063 
1064 	em_reset(ctx);
1065 
1066 	/* Initialize statistics */
1067 	em_update_stats_counters(adapter);
1068 	hw->mac.get_link_status = 1;
1069 	em_if_update_admin_status(ctx);
1070 	em_add_hw_stats(adapter);
1071 
1072 	/* Non-AMT based hardware can now take control from firmware */
1073 	if (adapter->has_manage && !adapter->has_amt)
1074 		em_get_hw_control(adapter);
1075 
1076 	INIT_DEBUGOUT("em_if_attach_post: end");
1077 
1078 	return (error);
1079 
1080 err_late:
1081 	em_release_hw_control(adapter);
1082 	em_free_pci_resources(ctx);
1083 	em_if_queues_free(ctx);
1084 	free(adapter->mta, M_DEVBUF);
1085 
1086 	return (error);
1087 }
1088 
1089 /*********************************************************************
1090  *  Device removal routine
1091  *
1092  *  The detach entry point is called when the driver is being removed.
1093  *  This routine stops the adapter and deallocates all the resources
1094  *  that were allocated for driver operation.
1095  *
1096  *  return 0 on success, positive on failure
1097  *********************************************************************/
1098 static int
1099 em_if_detach(if_ctx_t ctx)
1100 {
1101 	struct adapter	*adapter = iflib_get_softc(ctx);
1102 
1103 	INIT_DEBUGOUT("em_if_detach: begin");
1104 
1105 	e1000_phy_hw_reset(&adapter->hw);
1106 
1107 	em_release_manageability(adapter);
1108 	em_release_hw_control(adapter);
1109 	em_free_pci_resources(ctx);
1110 
1111 	return (0);
1112 }
1113 
1114 /*********************************************************************
1115  *
1116  *  Shutdown entry point
1117  *
1118  **********************************************************************/
1119 
1120 static int
1121 em_if_shutdown(if_ctx_t ctx)
1122 {
1123 	return em_if_suspend(ctx);
1124 }
1125 
1126 /*
1127  * Suspend/resume device methods.
1128  */
1129 static int
1130 em_if_suspend(if_ctx_t ctx)
1131 {
1132 	struct adapter *adapter = iflib_get_softc(ctx);
1133 
1134 	em_release_manageability(adapter);
1135 	em_release_hw_control(adapter);
1136 	em_enable_wakeup(ctx);
1137 	return (0);
1138 }
1139 
1140 static int
1141 em_if_resume(if_ctx_t ctx)
1142 {
1143 	struct adapter *adapter = iflib_get_softc(ctx);
1144 
1145 	if (adapter->hw.mac.type == e1000_pch2lan)
1146 		e1000_resume_workarounds_pchlan(&adapter->hw);
1147 	em_if_init(ctx);
1148 	em_init_manageability(adapter);
1149 
1150 	return(0);
1151 }
1152 
1153 static int
1154 em_if_mtu_set(if_ctx_t ctx, uint32_t mtu)
1155 {
1156 	int max_frame_size;
1157 	struct adapter *adapter = iflib_get_softc(ctx);
1158 	if_softc_ctx_t scctx = iflib_get_softc_ctx(ctx);
1159 
1160 	 IOCTL_DEBUGOUT("ioctl rcv'd: SIOCSIFMTU (Set Interface MTU)");
1161 
1162 	switch (adapter->hw.mac.type) {
1163 	case e1000_82571:
1164 	case e1000_82572:
1165 	case e1000_ich9lan:
1166 	case e1000_ich10lan:
1167 	case e1000_pch2lan:
1168 	case e1000_pch_lpt:
1169 	case e1000_pch_spt:
1170 	case e1000_pch_cnp:
1171 	case e1000_82574:
1172 	case e1000_82583:
1173 	case e1000_80003es2lan:
1174 		/* 9K Jumbo Frame size */
1175 		max_frame_size = 9234;
1176 		break;
1177 	case e1000_pchlan:
1178 		max_frame_size = 4096;
1179 		break;
1180 	case e1000_82542:
1181 	case e1000_ich8lan:
1182 		/* Adapters that do not support jumbo frames */
1183 		max_frame_size = ETHER_MAX_LEN;
1184 		break;
1185 	default:
1186 		if (adapter->hw.mac.type >= igb_mac_min)
1187 			max_frame_size = 9234;
1188 		else /* lem */
1189 			max_frame_size = MAX_JUMBO_FRAME_SIZE;
1190 	}
1191 	if (mtu > max_frame_size - ETHER_HDR_LEN - ETHER_CRC_LEN) {
1192 		return (EINVAL);
1193 	}
1194 
1195 	scctx->isc_max_frame_size = adapter->hw.mac.max_frame_size =
1196 	    mtu + ETHER_HDR_LEN + ETHER_CRC_LEN;
1197 	return (0);
1198 }
1199 
1200 /*********************************************************************
1201  *  Init entry point
1202  *
1203  *  This routine is used in two ways. It is used by the stack as
1204  *  init entry point in network interface structure. It is also used
1205  *  by the driver as a hw/sw initialization routine to get to a
1206  *  consistent state.
1207  *
1208  **********************************************************************/
1209 static void
1210 em_if_init(if_ctx_t ctx)
1211 {
1212 	struct adapter *adapter = iflib_get_softc(ctx);
1213 	if_softc_ctx_t scctx = adapter->shared;
1214 	struct ifnet *ifp = iflib_get_ifp(ctx);
1215 	struct em_tx_queue *tx_que;
1216 	int i;
1217 
1218 	INIT_DEBUGOUT("em_if_init: begin");
1219 
1220 	/* Get the latest mac address, User can use a LAA */
1221 	bcopy(if_getlladdr(ifp), adapter->hw.mac.addr,
1222 	    ETHER_ADDR_LEN);
1223 
1224 	/* Put the address into the Receive Address Array */
1225 	e1000_rar_set(&adapter->hw, adapter->hw.mac.addr, 0);
1226 
1227 	/*
1228 	 * With the 82571 adapter, RAR[0] may be overwritten
1229 	 * when the other port is reset, we make a duplicate
1230 	 * in RAR[14] for that eventuality, this assures
1231 	 * the interface continues to function.
1232 	 */
1233 	if (adapter->hw.mac.type == e1000_82571) {
1234 		e1000_set_laa_state_82571(&adapter->hw, TRUE);
1235 		e1000_rar_set(&adapter->hw, adapter->hw.mac.addr,
1236 		    E1000_RAR_ENTRIES - 1);
1237 	}
1238 
1239 
1240 	/* Initialize the hardware */
1241 	em_reset(ctx);
1242 	em_if_update_admin_status(ctx);
1243 
1244 	for (i = 0, tx_que = adapter->tx_queues; i < adapter->tx_num_queues; i++, tx_que++) {
1245 		struct tx_ring *txr = &tx_que->txr;
1246 
1247 		txr->tx_rs_cidx = txr->tx_rs_pidx;
1248 
1249 		/* Initialize the last processed descriptor to be the end of
1250 		 * the ring, rather than the start, so that we avoid an
1251 		 * off-by-one error when calculating how many descriptors are
1252 		 * done in the credits_update function.
1253 		 */
1254 		txr->tx_cidx_processed = scctx->isc_ntxd[0] - 1;
1255 	}
1256 
1257 	/* Setup VLAN support, basic and offload if available */
1258 	E1000_WRITE_REG(&adapter->hw, E1000_VET, ETHERTYPE_VLAN);
1259 
1260 	/* Clear bad data from Rx FIFOs */
1261 	if (adapter->hw.mac.type >= igb_mac_min)
1262 		e1000_rx_fifo_flush_82575(&adapter->hw);
1263 
1264 	/* Configure for OS presence */
1265 	em_init_manageability(adapter);
1266 
1267 	/* Prepare transmit descriptors and buffers */
1268 	em_initialize_transmit_unit(ctx);
1269 
1270 	/* Setup Multicast table */
1271 	em_if_multi_set(ctx);
1272 
1273 	/*
1274 	 * Figure out the desired mbuf
1275 	 * pool for doing jumbos
1276 	 */
1277 	if (adapter->hw.mac.max_frame_size <= 2048)
1278 		adapter->rx_mbuf_sz = MCLBYTES;
1279 #ifndef CONTIGMALLOC_WORKS
1280 	else
1281 		adapter->rx_mbuf_sz = MJUMPAGESIZE;
1282 #else
1283 	else if (adapter->hw.mac.max_frame_size <= 4096)
1284 		adapter->rx_mbuf_sz = MJUMPAGESIZE;
1285 	else
1286 		adapter->rx_mbuf_sz = MJUM9BYTES;
1287 #endif
1288 	em_initialize_receive_unit(ctx);
1289 
1290 	/* Use real VLAN Filter support? */
1291 	if (if_getcapenable(ifp) & IFCAP_VLAN_HWTAGGING) {
1292 		if (if_getcapenable(ifp) & IFCAP_VLAN_HWFILTER)
1293 			/* Use real VLAN Filter support */
1294 			em_setup_vlan_hw_support(adapter);
1295 		else {
1296 			u32 ctrl;
1297 			ctrl = E1000_READ_REG(&adapter->hw, E1000_CTRL);
1298 			ctrl |= E1000_CTRL_VME;
1299 			E1000_WRITE_REG(&adapter->hw, E1000_CTRL, ctrl);
1300 		}
1301 	}
1302 
1303 	/* Don't lose promiscuous settings */
1304 	em_if_set_promisc(ctx, IFF_PROMISC);
1305 	e1000_clear_hw_cntrs_base_generic(&adapter->hw);
1306 
1307 	/* MSI-X configuration for 82574 */
1308 	if (adapter->hw.mac.type == e1000_82574) {
1309 		int tmp = E1000_READ_REG(&adapter->hw, E1000_CTRL_EXT);
1310 
1311 		tmp |= E1000_CTRL_EXT_PBA_CLR;
1312 		E1000_WRITE_REG(&adapter->hw, E1000_CTRL_EXT, tmp);
1313 		/* Set the IVAR - interrupt vector routing. */
1314 		E1000_WRITE_REG(&adapter->hw, E1000_IVAR, adapter->ivars);
1315 	} else if (adapter->intr_type == IFLIB_INTR_MSIX) /* Set up queue routing */
1316 		igb_configure_queues(adapter);
1317 
1318 	/* this clears any pending interrupts */
1319 	E1000_READ_REG(&adapter->hw, E1000_ICR);
1320 	E1000_WRITE_REG(&adapter->hw, E1000_ICS, E1000_ICS_LSC);
1321 
1322 	/* AMT based hardware can now take control from firmware */
1323 	if (adapter->has_manage && adapter->has_amt)
1324 		em_get_hw_control(adapter);
1325 
1326 	/* Set Energy Efficient Ethernet */
1327 	if (adapter->hw.mac.type >= igb_mac_min &&
1328 	    adapter->hw.phy.media_type == e1000_media_type_copper) {
1329 		if (adapter->hw.mac.type == e1000_i354)
1330 			e1000_set_eee_i354(&adapter->hw, TRUE, TRUE);
1331 		else
1332 			e1000_set_eee_i350(&adapter->hw, TRUE, TRUE);
1333 	}
1334 }
1335 
1336 /*********************************************************************
1337  *
1338  *  Fast Legacy/MSI Combined Interrupt Service routine
1339  *
1340  *********************************************************************/
1341 int
1342 em_intr(void *arg)
1343 {
1344 	struct adapter *adapter = arg;
1345 	if_ctx_t ctx = adapter->ctx;
1346 	u32 reg_icr;
1347 
1348 	reg_icr = E1000_READ_REG(&adapter->hw, E1000_ICR);
1349 
1350 	if (adapter->intr_type != IFLIB_INTR_LEGACY)
1351 		goto skip_stray;
1352 	/* Hot eject? */
1353 	if (reg_icr == 0xffffffff)
1354 		return FILTER_STRAY;
1355 
1356 	/* Definitely not our interrupt. */
1357 	if (reg_icr == 0x0)
1358 		return FILTER_STRAY;
1359 
1360 	/*
1361 	 * Starting with the 82571 chip, bit 31 should be used to
1362 	 * determine whether the interrupt belongs to us.
1363 	 */
1364 	if (adapter->hw.mac.type >= e1000_82571 &&
1365 	    (reg_icr & E1000_ICR_INT_ASSERTED) == 0)
1366 		return FILTER_STRAY;
1367 
1368 skip_stray:
1369 	/* Link status change */
1370 	if (reg_icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) {
1371 		adapter->hw.mac.get_link_status = 1;
1372 		iflib_admin_intr_deferred(ctx);
1373 	}
1374 
1375 	if (reg_icr & E1000_ICR_RXO)
1376 		adapter->rx_overruns++;
1377 
1378 	return (FILTER_SCHEDULE_THREAD);
1379 }
1380 
1381 static void
1382 igb_rx_enable_queue(struct adapter *adapter, struct em_rx_queue *rxq)
1383 {
1384 	E1000_WRITE_REG(&adapter->hw, E1000_EIMS, rxq->eims);
1385 }
1386 
1387 static void
1388 em_rx_enable_queue(struct adapter *adapter, struct em_rx_queue *rxq)
1389 {
1390 	E1000_WRITE_REG(&adapter->hw, E1000_IMS, rxq->eims);
1391 }
1392 
1393 static void
1394 igb_tx_enable_queue(struct adapter *adapter, struct em_tx_queue *txq)
1395 {
1396 	E1000_WRITE_REG(&adapter->hw, E1000_EIMS, txq->eims);
1397 }
1398 
1399 static void
1400 em_tx_enable_queue(struct adapter *adapter, struct em_tx_queue *txq)
1401 {
1402 	E1000_WRITE_REG(&adapter->hw, E1000_IMS, txq->eims);
1403 }
1404 
1405 static int
1406 em_if_rx_queue_intr_enable(if_ctx_t ctx, uint16_t rxqid)
1407 {
1408 	struct adapter *adapter = iflib_get_softc(ctx);
1409 	struct em_rx_queue *rxq = &adapter->rx_queues[rxqid];
1410 
1411 	if (adapter->hw.mac.type >= igb_mac_min)
1412 		igb_rx_enable_queue(adapter, rxq);
1413 	else
1414 		em_rx_enable_queue(adapter, rxq);
1415 	return (0);
1416 }
1417 
1418 static int
1419 em_if_tx_queue_intr_enable(if_ctx_t ctx, uint16_t txqid)
1420 {
1421 	struct adapter *adapter = iflib_get_softc(ctx);
1422 	struct em_tx_queue *txq = &adapter->tx_queues[txqid];
1423 
1424 	if (adapter->hw.mac.type >= igb_mac_min)
1425 		igb_tx_enable_queue(adapter, txq);
1426 	else
1427 		em_tx_enable_queue(adapter, txq);
1428 	return (0);
1429 }
1430 
1431 /*********************************************************************
1432  *
1433  *  MSI-X RX Interrupt Service routine
1434  *
1435  **********************************************************************/
1436 static int
1437 em_msix_que(void *arg)
1438 {
1439 	struct em_rx_queue *que = arg;
1440 
1441 	++que->irqs;
1442 
1443 	return (FILTER_SCHEDULE_THREAD);
1444 }
1445 
1446 /*********************************************************************
1447  *
1448  *  MSI-X Link Fast Interrupt Service routine
1449  *
1450  **********************************************************************/
1451 static int
1452 em_msix_link(void *arg)
1453 {
1454 	struct adapter *adapter = arg;
1455 	u32 reg_icr;
1456 
1457 	++adapter->link_irq;
1458 	MPASS(adapter->hw.back != NULL);
1459 	reg_icr = E1000_READ_REG(&adapter->hw, E1000_ICR);
1460 
1461 	if (reg_icr & E1000_ICR_RXO)
1462 		adapter->rx_overruns++;
1463 
1464 	if (reg_icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) {
1465 		em_handle_link(adapter->ctx);
1466 	} else {
1467 		E1000_WRITE_REG(&adapter->hw, E1000_IMS,
1468 				EM_MSIX_LINK | E1000_IMS_LSC);
1469 		if (adapter->hw.mac.type >= igb_mac_min)
1470 			E1000_WRITE_REG(&adapter->hw, E1000_EIMS, adapter->link_mask);
1471 	}
1472 
1473 	/*
1474 	 * Because we must read the ICR for this interrupt
1475 	 * it may clear other causes using autoclear, for
1476 	 * this reason we simply create a soft interrupt
1477 	 * for all these vectors.
1478 	 */
1479 	if (reg_icr && adapter->hw.mac.type < igb_mac_min) {
1480 		E1000_WRITE_REG(&adapter->hw,
1481 			E1000_ICS, adapter->ims);
1482 	}
1483 
1484 	return (FILTER_HANDLED);
1485 }
1486 
1487 static void
1488 em_handle_link(void *context)
1489 {
1490 	if_ctx_t ctx = context;
1491 	struct adapter *adapter = iflib_get_softc(ctx);
1492 
1493 	adapter->hw.mac.get_link_status = 1;
1494 	iflib_admin_intr_deferred(ctx);
1495 }
1496 
1497 
1498 /*********************************************************************
1499  *
1500  *  Media Ioctl callback
1501  *
1502  *  This routine is called whenever the user queries the status of
1503  *  the interface using ifconfig.
1504  *
1505  **********************************************************************/
1506 static void
1507 em_if_media_status(if_ctx_t ctx, struct ifmediareq *ifmr)
1508 {
1509 	struct adapter *adapter = iflib_get_softc(ctx);
1510 	u_char fiber_type = IFM_1000_SX;
1511 
1512 	INIT_DEBUGOUT("em_if_media_status: begin");
1513 
1514 	iflib_admin_intr_deferred(ctx);
1515 
1516 	ifmr->ifm_status = IFM_AVALID;
1517 	ifmr->ifm_active = IFM_ETHER;
1518 
1519 	if (!adapter->link_active) {
1520 		return;
1521 	}
1522 
1523 	ifmr->ifm_status |= IFM_ACTIVE;
1524 
1525 	if ((adapter->hw.phy.media_type == e1000_media_type_fiber) ||
1526 	    (adapter->hw.phy.media_type == e1000_media_type_internal_serdes)) {
1527 		if (adapter->hw.mac.type == e1000_82545)
1528 			fiber_type = IFM_1000_LX;
1529 		ifmr->ifm_active |= fiber_type | IFM_FDX;
1530 	} else {
1531 		switch (adapter->link_speed) {
1532 		case 10:
1533 			ifmr->ifm_active |= IFM_10_T;
1534 			break;
1535 		case 100:
1536 			ifmr->ifm_active |= IFM_100_TX;
1537 			break;
1538 		case 1000:
1539 			ifmr->ifm_active |= IFM_1000_T;
1540 			break;
1541 		}
1542 		if (adapter->link_duplex == FULL_DUPLEX)
1543 			ifmr->ifm_active |= IFM_FDX;
1544 		else
1545 			ifmr->ifm_active |= IFM_HDX;
1546 	}
1547 }
1548 
1549 /*********************************************************************
1550  *
1551  *  Media Ioctl callback
1552  *
1553  *  This routine is called when the user changes speed/duplex using
1554  *  media/mediopt option with ifconfig.
1555  *
1556  **********************************************************************/
1557 static int
1558 em_if_media_change(if_ctx_t ctx)
1559 {
1560 	struct adapter *adapter = iflib_get_softc(ctx);
1561 	struct ifmedia *ifm = iflib_get_media(ctx);
1562 
1563 	INIT_DEBUGOUT("em_if_media_change: begin");
1564 
1565 	if (IFM_TYPE(ifm->ifm_media) != IFM_ETHER)
1566 		return (EINVAL);
1567 
1568 	switch (IFM_SUBTYPE(ifm->ifm_media)) {
1569 	case IFM_AUTO:
1570 		adapter->hw.mac.autoneg = DO_AUTO_NEG;
1571 		adapter->hw.phy.autoneg_advertised = AUTONEG_ADV_DEFAULT;
1572 		break;
1573 	case IFM_1000_LX:
1574 	case IFM_1000_SX:
1575 	case IFM_1000_T:
1576 		adapter->hw.mac.autoneg = DO_AUTO_NEG;
1577 		adapter->hw.phy.autoneg_advertised = ADVERTISE_1000_FULL;
1578 		break;
1579 	case IFM_100_TX:
1580 		adapter->hw.mac.autoneg = FALSE;
1581 		adapter->hw.phy.autoneg_advertised = 0;
1582 		if ((ifm->ifm_media & IFM_GMASK) == IFM_FDX)
1583 			adapter->hw.mac.forced_speed_duplex = ADVERTISE_100_FULL;
1584 		else
1585 			adapter->hw.mac.forced_speed_duplex = ADVERTISE_100_HALF;
1586 		break;
1587 	case IFM_10_T:
1588 		adapter->hw.mac.autoneg = FALSE;
1589 		adapter->hw.phy.autoneg_advertised = 0;
1590 		if ((ifm->ifm_media & IFM_GMASK) == IFM_FDX)
1591 			adapter->hw.mac.forced_speed_duplex = ADVERTISE_10_FULL;
1592 		else
1593 			adapter->hw.mac.forced_speed_duplex = ADVERTISE_10_HALF;
1594 		break;
1595 	default:
1596 		device_printf(adapter->dev, "Unsupported media type\n");
1597 	}
1598 
1599 	em_if_init(ctx);
1600 
1601 	return (0);
1602 }
1603 
1604 static int
1605 em_if_set_promisc(if_ctx_t ctx, int flags)
1606 {
1607 	struct adapter *adapter = iflib_get_softc(ctx);
1608 	u32 reg_rctl;
1609 
1610 	em_disable_promisc(ctx);
1611 
1612 	reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
1613 
1614 	if (flags & IFF_PROMISC) {
1615 		reg_rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
1616 		/* Turn this on if you want to see bad packets */
1617 		if (em_debug_sbp)
1618 			reg_rctl |= E1000_RCTL_SBP;
1619 		E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1620 	} else if (flags & IFF_ALLMULTI) {
1621 		reg_rctl |= E1000_RCTL_MPE;
1622 		reg_rctl &= ~E1000_RCTL_UPE;
1623 		E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1624 	}
1625 	return (0);
1626 }
1627 
1628 static void
1629 em_disable_promisc(if_ctx_t ctx)
1630 {
1631 	struct adapter *adapter = iflib_get_softc(ctx);
1632 	struct ifnet *ifp = iflib_get_ifp(ctx);
1633 	u32 reg_rctl;
1634 	int mcnt = 0;
1635 
1636 	reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
1637 	reg_rctl &= (~E1000_RCTL_UPE);
1638 	if (if_getflags(ifp) & IFF_ALLMULTI)
1639 		mcnt = MAX_NUM_MULTICAST_ADDRESSES;
1640 	else
1641 		mcnt = if_multiaddr_count(ifp, MAX_NUM_MULTICAST_ADDRESSES);
1642 	/* Don't disable if in MAX groups */
1643 	if (mcnt < MAX_NUM_MULTICAST_ADDRESSES)
1644 		reg_rctl &=  (~E1000_RCTL_MPE);
1645 	reg_rctl &=  (~E1000_RCTL_SBP);
1646 	E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1647 }
1648 
1649 
1650 /*********************************************************************
1651  *  Multicast Update
1652  *
1653  *  This routine is called whenever multicast address list is updated.
1654  *
1655  **********************************************************************/
1656 
1657 static void
1658 em_if_multi_set(if_ctx_t ctx)
1659 {
1660 	struct adapter *adapter = iflib_get_softc(ctx);
1661 	struct ifnet *ifp = iflib_get_ifp(ctx);
1662 	u32 reg_rctl = 0;
1663 	u8  *mta; /* Multicast array memory */
1664 	int mcnt = 0;
1665 
1666 	IOCTL_DEBUGOUT("em_set_multi: begin");
1667 
1668 	mta = adapter->mta;
1669 	bzero(mta, sizeof(u8) * ETH_ADDR_LEN * MAX_NUM_MULTICAST_ADDRESSES);
1670 
1671 	if (adapter->hw.mac.type == e1000_82542 &&
1672 	    adapter->hw.revision_id == E1000_REVISION_2) {
1673 		reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
1674 		if (adapter->hw.bus.pci_cmd_word & CMD_MEM_WRT_INVALIDATE)
1675 			e1000_pci_clear_mwi(&adapter->hw);
1676 		reg_rctl |= E1000_RCTL_RST;
1677 		E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1678 		msec_delay(5);
1679 	}
1680 
1681 	if_multiaddr_array(ifp, mta, &mcnt, MAX_NUM_MULTICAST_ADDRESSES);
1682 
1683 	if (mcnt >= MAX_NUM_MULTICAST_ADDRESSES) {
1684 		reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
1685 		reg_rctl |= E1000_RCTL_MPE;
1686 		E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1687 	} else
1688 		e1000_update_mc_addr_list(&adapter->hw, mta, mcnt);
1689 
1690 	if (adapter->hw.mac.type == e1000_82542 &&
1691 	    adapter->hw.revision_id == E1000_REVISION_2) {
1692 		reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
1693 		reg_rctl &= ~E1000_RCTL_RST;
1694 		E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1695 		msec_delay(5);
1696 		if (adapter->hw.bus.pci_cmd_word & CMD_MEM_WRT_INVALIDATE)
1697 			e1000_pci_set_mwi(&adapter->hw);
1698 	}
1699 }
1700 
1701 /*********************************************************************
1702  *  Timer routine
1703  *
1704  *  This routine schedules em_if_update_admin_status() to check for
1705  *  link status and to gather statistics as well as to perform some
1706  *  controller-specific hardware patting.
1707  *
1708  **********************************************************************/
1709 static void
1710 em_if_timer(if_ctx_t ctx, uint16_t qid)
1711 {
1712 
1713 	if (qid != 0)
1714 		return;
1715 
1716 	iflib_admin_intr_deferred(ctx);
1717 }
1718 
1719 static void
1720 em_if_update_admin_status(if_ctx_t ctx)
1721 {
1722 	struct adapter *adapter = iflib_get_softc(ctx);
1723 	struct e1000_hw *hw = &adapter->hw;
1724 	device_t dev = iflib_get_dev(ctx);
1725 	u32 link_check, thstat, ctrl;
1726 
1727 	link_check = thstat = ctrl = 0;
1728 	/* Get the cached link value or read phy for real */
1729 	switch (hw->phy.media_type) {
1730 	case e1000_media_type_copper:
1731 		if (hw->mac.get_link_status) {
1732 			if (hw->mac.type == e1000_pch_spt)
1733 				msec_delay(50);
1734 			/* Do the work to read phy */
1735 			e1000_check_for_link(hw);
1736 			link_check = !hw->mac.get_link_status;
1737 			if (link_check) /* ESB2 fix */
1738 				e1000_cfg_on_link_up(hw);
1739 		} else {
1740 			link_check = TRUE;
1741 		}
1742 		break;
1743 	case e1000_media_type_fiber:
1744 		e1000_check_for_link(hw);
1745 		link_check = (E1000_READ_REG(hw, E1000_STATUS) &
1746 			    E1000_STATUS_LU);
1747 		break;
1748 	case e1000_media_type_internal_serdes:
1749 		e1000_check_for_link(hw);
1750 		link_check = adapter->hw.mac.serdes_has_link;
1751 		break;
1752 	/* VF device is type_unknown */
1753 	case e1000_media_type_unknown:
1754 		e1000_check_for_link(hw);
1755 		link_check = !hw->mac.get_link_status;
1756 		/* FALLTHROUGH */
1757 	default:
1758 		break;
1759 	}
1760 
1761 	/* Check for thermal downshift or shutdown */
1762 	if (hw->mac.type == e1000_i350) {
1763 		thstat = E1000_READ_REG(hw, E1000_THSTAT);
1764 		ctrl = E1000_READ_REG(hw, E1000_CTRL_EXT);
1765 	}
1766 
1767 	/* Now check for a transition */
1768 	if (link_check && (adapter->link_active == 0)) {
1769 		e1000_get_speed_and_duplex(hw, &adapter->link_speed,
1770 		    &adapter->link_duplex);
1771 		/* Check if we must disable SPEED_MODE bit on PCI-E */
1772 		if ((adapter->link_speed != SPEED_1000) &&
1773 		    ((hw->mac.type == e1000_82571) ||
1774 		    (hw->mac.type == e1000_82572))) {
1775 			int tarc0;
1776 			tarc0 = E1000_READ_REG(hw, E1000_TARC(0));
1777 			tarc0 &= ~TARC_SPEED_MODE_BIT;
1778 			E1000_WRITE_REG(hw, E1000_TARC(0), tarc0);
1779 		}
1780 		if (bootverbose)
1781 			device_printf(dev, "Link is up %d Mbps %s\n",
1782 			    adapter->link_speed,
1783 			    ((adapter->link_duplex == FULL_DUPLEX) ?
1784 			    "Full Duplex" : "Half Duplex"));
1785 		adapter->link_active = 1;
1786 		adapter->smartspeed = 0;
1787 		if ((ctrl & E1000_CTRL_EXT_LINK_MODE_MASK) ==
1788 		    E1000_CTRL_EXT_LINK_MODE_GMII &&
1789 		    (thstat & E1000_THSTAT_LINK_THROTTLE))
1790 			device_printf(dev, "Link: thermal downshift\n");
1791 		/* Delay Link Up for Phy update */
1792 		if (((hw->mac.type == e1000_i210) ||
1793 		    (hw->mac.type == e1000_i211)) &&
1794 		    (hw->phy.id == I210_I_PHY_ID))
1795 			msec_delay(I210_LINK_DELAY);
1796 		/* Reset if the media type changed. */
1797 		if ((hw->dev_spec._82575.media_changed) &&
1798 			(adapter->hw.mac.type >= igb_mac_min)) {
1799 			hw->dev_spec._82575.media_changed = false;
1800 			adapter->flags |= IGB_MEDIA_RESET;
1801 			em_reset(ctx);
1802 		}
1803 		iflib_link_state_change(ctx, LINK_STATE_UP,
1804 		    IF_Mbps(adapter->link_speed));
1805 	} else if (!link_check && (adapter->link_active == 1)) {
1806 		adapter->link_speed = 0;
1807 		adapter->link_duplex = 0;
1808 		adapter->link_active = 0;
1809 		iflib_link_state_change(ctx, LINK_STATE_DOWN, 0);
1810 	}
1811 	em_update_stats_counters(adapter);
1812 
1813 	/* Reset LAA into RAR[0] on 82571 */
1814 	if ((adapter->hw.mac.type == e1000_82571) &&
1815 	    e1000_get_laa_state_82571(&adapter->hw))
1816 		e1000_rar_set(&adapter->hw, adapter->hw.mac.addr, 0);
1817 
1818 	if (adapter->hw.mac.type < em_mac_min)
1819 		lem_smartspeed(adapter);
1820 
1821 	E1000_WRITE_REG(&adapter->hw, E1000_IMS, EM_MSIX_LINK | E1000_IMS_LSC);
1822 }
1823 
1824 static void
1825 em_if_watchdog_reset(if_ctx_t ctx)
1826 {
1827 	struct adapter *adapter = iflib_get_softc(ctx);
1828 
1829 	/*
1830 	 * Just count the event; iflib(4) will already trigger a
1831 	 * sufficient reset of the controller.
1832 	 */
1833 	adapter->watchdog_events++;
1834 }
1835 
1836 /*********************************************************************
1837  *
1838  *  This routine disables all traffic on the adapter by issuing a
1839  *  global reset on the MAC.
1840  *
1841  **********************************************************************/
1842 static void
1843 em_if_stop(if_ctx_t ctx)
1844 {
1845 	struct adapter *adapter = iflib_get_softc(ctx);
1846 
1847 	INIT_DEBUGOUT("em_if_stop: begin");
1848 
1849 	e1000_reset_hw(&adapter->hw);
1850 	if (adapter->hw.mac.type >= e1000_82544)
1851 		E1000_WRITE_REG(&adapter->hw, E1000_WUFC, 0);
1852 
1853 	e1000_led_off(&adapter->hw);
1854 	e1000_cleanup_led(&adapter->hw);
1855 }
1856 
1857 /*********************************************************************
1858  *
1859  *  Determine hardware revision.
1860  *
1861  **********************************************************************/
1862 static void
1863 em_identify_hardware(if_ctx_t ctx)
1864 {
1865 	device_t dev = iflib_get_dev(ctx);
1866 	struct adapter *adapter = iflib_get_softc(ctx);
1867 
1868 	/* Make sure our PCI config space has the necessary stuff set */
1869 	adapter->hw.bus.pci_cmd_word = pci_read_config(dev, PCIR_COMMAND, 2);
1870 
1871 	/* Save off the information about this board */
1872 	adapter->hw.vendor_id = pci_get_vendor(dev);
1873 	adapter->hw.device_id = pci_get_device(dev);
1874 	adapter->hw.revision_id = pci_read_config(dev, PCIR_REVID, 1);
1875 	adapter->hw.subsystem_vendor_id =
1876 	    pci_read_config(dev, PCIR_SUBVEND_0, 2);
1877 	adapter->hw.subsystem_device_id =
1878 	    pci_read_config(dev, PCIR_SUBDEV_0, 2);
1879 
1880 	/* Do Shared Code Init and Setup */
1881 	if (e1000_set_mac_type(&adapter->hw)) {
1882 		device_printf(dev, "Setup init failure\n");
1883 		return;
1884 	}
1885 }
1886 
1887 static int
1888 em_allocate_pci_resources(if_ctx_t ctx)
1889 {
1890 	struct adapter *adapter = iflib_get_softc(ctx);
1891 	device_t dev = iflib_get_dev(ctx);
1892 	int rid, val;
1893 
1894 	rid = PCIR_BAR(0);
1895 	adapter->memory = bus_alloc_resource_any(dev, SYS_RES_MEMORY,
1896 	    &rid, RF_ACTIVE);
1897 	if (adapter->memory == NULL) {
1898 		device_printf(dev, "Unable to allocate bus resource: memory\n");
1899 		return (ENXIO);
1900 	}
1901 	adapter->osdep.mem_bus_space_tag = rman_get_bustag(adapter->memory);
1902 	adapter->osdep.mem_bus_space_handle =
1903 	    rman_get_bushandle(adapter->memory);
1904 	adapter->hw.hw_addr = (u8 *)&adapter->osdep.mem_bus_space_handle;
1905 
1906 	/* Only older adapters use IO mapping */
1907 	if (adapter->hw.mac.type < em_mac_min &&
1908 	    adapter->hw.mac.type > e1000_82543) {
1909 		/* Figure our where our IO BAR is ? */
1910 		for (rid = PCIR_BAR(0); rid < PCIR_CIS;) {
1911 			val = pci_read_config(dev, rid, 4);
1912 			if (EM_BAR_TYPE(val) == EM_BAR_TYPE_IO) {
1913 				break;
1914 			}
1915 			rid += 4;
1916 			/* check for 64bit BAR */
1917 			if (EM_BAR_MEM_TYPE(val) == EM_BAR_MEM_TYPE_64BIT)
1918 				rid += 4;
1919 		}
1920 		if (rid >= PCIR_CIS) {
1921 			device_printf(dev, "Unable to locate IO BAR\n");
1922 			return (ENXIO);
1923 		}
1924 		adapter->ioport = bus_alloc_resource_any(dev, SYS_RES_IOPORT,
1925 		    &rid, RF_ACTIVE);
1926 		if (adapter->ioport == NULL) {
1927 			device_printf(dev, "Unable to allocate bus resource: "
1928 			    "ioport\n");
1929 			return (ENXIO);
1930 		}
1931 		adapter->hw.io_base = 0;
1932 		adapter->osdep.io_bus_space_tag =
1933 		    rman_get_bustag(adapter->ioport);
1934 		adapter->osdep.io_bus_space_handle =
1935 		    rman_get_bushandle(adapter->ioport);
1936 	}
1937 
1938 	adapter->hw.back = &adapter->osdep;
1939 
1940 	return (0);
1941 }
1942 
1943 /*********************************************************************
1944  *
1945  *  Set up the MSI-X Interrupt handlers
1946  *
1947  **********************************************************************/
1948 static int
1949 em_if_msix_intr_assign(if_ctx_t ctx, int msix)
1950 {
1951 	struct adapter *adapter = iflib_get_softc(ctx);
1952 	struct em_rx_queue *rx_que = adapter->rx_queues;
1953 	struct em_tx_queue *tx_que = adapter->tx_queues;
1954 	int error, rid, i, vector = 0, rx_vectors;
1955 	char buf[16];
1956 
1957 	/* First set up ring resources */
1958 	for (i = 0; i < adapter->rx_num_queues; i++, rx_que++, vector++) {
1959 		rid = vector + 1;
1960 		snprintf(buf, sizeof(buf), "rxq%d", i);
1961 		error = iflib_irq_alloc_generic(ctx, &rx_que->que_irq, rid, IFLIB_INTR_RXTX, em_msix_que, rx_que, rx_que->me, buf);
1962 		if (error) {
1963 			device_printf(iflib_get_dev(ctx), "Failed to allocate que int %d err: %d", i, error);
1964 			adapter->rx_num_queues = i + 1;
1965 			goto fail;
1966 		}
1967 
1968 		rx_que->msix =  vector;
1969 
1970 		/*
1971 		 * Set the bit to enable interrupt
1972 		 * in E1000_IMS -- bits 20 and 21
1973 		 * are for RX0 and RX1, note this has
1974 		 * NOTHING to do with the MSI-X vector
1975 		 */
1976 		if (adapter->hw.mac.type == e1000_82574) {
1977 			rx_que->eims = 1 << (20 + i);
1978 			adapter->ims |= rx_que->eims;
1979 			adapter->ivars |= (8 | rx_que->msix) << (i * 4);
1980 		} else if (adapter->hw.mac.type == e1000_82575)
1981 			rx_que->eims = E1000_EICR_TX_QUEUE0 << vector;
1982 		else
1983 			rx_que->eims = 1 << vector;
1984 	}
1985 	rx_vectors = vector;
1986 
1987 	vector = 0;
1988 	for (i = 0; i < adapter->tx_num_queues; i++, tx_que++, vector++) {
1989 		snprintf(buf, sizeof(buf), "txq%d", i);
1990 		tx_que = &adapter->tx_queues[i];
1991 		iflib_softirq_alloc_generic(ctx,
1992 		    &adapter->rx_queues[i % adapter->rx_num_queues].que_irq,
1993 		    IFLIB_INTR_TX, tx_que, tx_que->me, buf);
1994 
1995 		tx_que->msix = (vector % adapter->rx_num_queues);
1996 
1997 		/*
1998 		 * Set the bit to enable interrupt
1999 		 * in E1000_IMS -- bits 22 and 23
2000 		 * are for TX0 and TX1, note this has
2001 		 * NOTHING to do with the MSI-X vector
2002 		 */
2003 		if (adapter->hw.mac.type == e1000_82574) {
2004 			tx_que->eims = 1 << (22 + i);
2005 			adapter->ims |= tx_que->eims;
2006 			adapter->ivars |= (8 | tx_que->msix) << (8 + (i * 4));
2007 		} else if (adapter->hw.mac.type == e1000_82575) {
2008 			tx_que->eims = E1000_EICR_TX_QUEUE0 << i;
2009 		} else {
2010 			tx_que->eims = 1 << i;
2011 		}
2012 	}
2013 
2014 	/* Link interrupt */
2015 	rid = rx_vectors + 1;
2016 	error = iflib_irq_alloc_generic(ctx, &adapter->irq, rid, IFLIB_INTR_ADMIN, em_msix_link, adapter, 0, "aq");
2017 
2018 	if (error) {
2019 		device_printf(iflib_get_dev(ctx), "Failed to register admin handler");
2020 		goto fail;
2021 	}
2022 	adapter->linkvec = rx_vectors;
2023 	if (adapter->hw.mac.type < igb_mac_min) {
2024 		adapter->ivars |=  (8 | rx_vectors) << 16;
2025 		adapter->ivars |= 0x80000000;
2026 	}
2027 	return (0);
2028 fail:
2029 	iflib_irq_free(ctx, &adapter->irq);
2030 	rx_que = adapter->rx_queues;
2031 	for (int i = 0; i < adapter->rx_num_queues; i++, rx_que++)
2032 		iflib_irq_free(ctx, &rx_que->que_irq);
2033 	return (error);
2034 }
2035 
2036 static void
2037 igb_configure_queues(struct adapter *adapter)
2038 {
2039 	struct e1000_hw *hw = &adapter->hw;
2040 	struct em_rx_queue *rx_que;
2041 	struct em_tx_queue *tx_que;
2042 	u32 tmp, ivar = 0, newitr = 0;
2043 
2044 	/* First turn on RSS capability */
2045 	if (adapter->hw.mac.type != e1000_82575)
2046 		E1000_WRITE_REG(hw, E1000_GPIE,
2047 		    E1000_GPIE_MSIX_MODE | E1000_GPIE_EIAME |
2048 		    E1000_GPIE_PBA | E1000_GPIE_NSICR);
2049 
2050 	/* Turn on MSI-X */
2051 	switch (adapter->hw.mac.type) {
2052 	case e1000_82580:
2053 	case e1000_i350:
2054 	case e1000_i354:
2055 	case e1000_i210:
2056 	case e1000_i211:
2057 	case e1000_vfadapt:
2058 	case e1000_vfadapt_i350:
2059 		/* RX entries */
2060 		for (int i = 0; i < adapter->rx_num_queues; i++) {
2061 			u32 index = i >> 1;
2062 			ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2063 			rx_que = &adapter->rx_queues[i];
2064 			if (i & 1) {
2065 				ivar &= 0xFF00FFFF;
2066 				ivar |= (rx_que->msix | E1000_IVAR_VALID) << 16;
2067 			} else {
2068 				ivar &= 0xFFFFFF00;
2069 				ivar |= rx_que->msix | E1000_IVAR_VALID;
2070 			}
2071 			E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2072 		}
2073 		/* TX entries */
2074 		for (int i = 0; i < adapter->tx_num_queues; i++) {
2075 			u32 index = i >> 1;
2076 			ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2077 			tx_que = &adapter->tx_queues[i];
2078 			if (i & 1) {
2079 				ivar &= 0x00FFFFFF;
2080 				ivar |= (tx_que->msix | E1000_IVAR_VALID) << 24;
2081 			} else {
2082 				ivar &= 0xFFFF00FF;
2083 				ivar |= (tx_que->msix | E1000_IVAR_VALID) << 8;
2084 			}
2085 			E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2086 			adapter->que_mask |= tx_que->eims;
2087 		}
2088 
2089 		/* And for the link interrupt */
2090 		ivar = (adapter->linkvec | E1000_IVAR_VALID) << 8;
2091 		adapter->link_mask = 1 << adapter->linkvec;
2092 		E1000_WRITE_REG(hw, E1000_IVAR_MISC, ivar);
2093 		break;
2094 	case e1000_82576:
2095 		/* RX entries */
2096 		for (int i = 0; i < adapter->rx_num_queues; i++) {
2097 			u32 index = i & 0x7; /* Each IVAR has two entries */
2098 			ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2099 			rx_que = &adapter->rx_queues[i];
2100 			if (i < 8) {
2101 				ivar &= 0xFFFFFF00;
2102 				ivar |= rx_que->msix | E1000_IVAR_VALID;
2103 			} else {
2104 				ivar &= 0xFF00FFFF;
2105 				ivar |= (rx_que->msix | E1000_IVAR_VALID) << 16;
2106 			}
2107 			E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2108 			adapter->que_mask |= rx_que->eims;
2109 		}
2110 		/* TX entries */
2111 		for (int i = 0; i < adapter->tx_num_queues; i++) {
2112 			u32 index = i & 0x7; /* Each IVAR has two entries */
2113 			ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2114 			tx_que = &adapter->tx_queues[i];
2115 			if (i < 8) {
2116 				ivar &= 0xFFFF00FF;
2117 				ivar |= (tx_que->msix | E1000_IVAR_VALID) << 8;
2118 			} else {
2119 				ivar &= 0x00FFFFFF;
2120 				ivar |= (tx_que->msix | E1000_IVAR_VALID) << 24;
2121 			}
2122 			E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2123 			adapter->que_mask |= tx_que->eims;
2124 		}
2125 
2126 		/* And for the link interrupt */
2127 		ivar = (adapter->linkvec | E1000_IVAR_VALID) << 8;
2128 		adapter->link_mask = 1 << adapter->linkvec;
2129 		E1000_WRITE_REG(hw, E1000_IVAR_MISC, ivar);
2130 		break;
2131 
2132 	case e1000_82575:
2133 		/* enable MSI-X support*/
2134 		tmp = E1000_READ_REG(hw, E1000_CTRL_EXT);
2135 		tmp |= E1000_CTRL_EXT_PBA_CLR;
2136 		/* Auto-Mask interrupts upon ICR read. */
2137 		tmp |= E1000_CTRL_EXT_EIAME;
2138 		tmp |= E1000_CTRL_EXT_IRCA;
2139 		E1000_WRITE_REG(hw, E1000_CTRL_EXT, tmp);
2140 
2141 		/* Queues */
2142 		for (int i = 0; i < adapter->rx_num_queues; i++) {
2143 			rx_que = &adapter->rx_queues[i];
2144 			tmp = E1000_EICR_RX_QUEUE0 << i;
2145 			tmp |= E1000_EICR_TX_QUEUE0 << i;
2146 			rx_que->eims = tmp;
2147 			E1000_WRITE_REG_ARRAY(hw, E1000_MSIXBM(0),
2148 			    i, rx_que->eims);
2149 			adapter->que_mask |= rx_que->eims;
2150 		}
2151 
2152 		/* Link */
2153 		E1000_WRITE_REG(hw, E1000_MSIXBM(adapter->linkvec),
2154 		    E1000_EIMS_OTHER);
2155 		adapter->link_mask |= E1000_EIMS_OTHER;
2156 	default:
2157 		break;
2158 	}
2159 
2160 	/* Set the starting interrupt rate */
2161 	if (em_max_interrupt_rate > 0)
2162 		newitr = (4000000 / em_max_interrupt_rate) & 0x7FFC;
2163 
2164 	if (hw->mac.type == e1000_82575)
2165 		newitr |= newitr << 16;
2166 	else
2167 		newitr |= E1000_EITR_CNT_IGNR;
2168 
2169 	for (int i = 0; i < adapter->rx_num_queues; i++) {
2170 		rx_que = &adapter->rx_queues[i];
2171 		E1000_WRITE_REG(hw, E1000_EITR(rx_que->msix), newitr);
2172 	}
2173 
2174 	return;
2175 }
2176 
2177 static void
2178 em_free_pci_resources(if_ctx_t ctx)
2179 {
2180 	struct adapter *adapter = iflib_get_softc(ctx);
2181 	struct em_rx_queue *que = adapter->rx_queues;
2182 	device_t dev = iflib_get_dev(ctx);
2183 
2184 	/* Release all MSI-X queue resources */
2185 	if (adapter->intr_type == IFLIB_INTR_MSIX)
2186 		iflib_irq_free(ctx, &adapter->irq);
2187 
2188 	for (int i = 0; i < adapter->rx_num_queues; i++, que++) {
2189 		iflib_irq_free(ctx, &que->que_irq);
2190 	}
2191 
2192 	if (adapter->memory != NULL) {
2193 		bus_release_resource(dev, SYS_RES_MEMORY,
2194 		    rman_get_rid(adapter->memory), adapter->memory);
2195 		adapter->memory = NULL;
2196 	}
2197 
2198 	if (adapter->flash != NULL) {
2199 		bus_release_resource(dev, SYS_RES_MEMORY,
2200 		    rman_get_rid(adapter->flash), adapter->flash);
2201 		adapter->flash = NULL;
2202 	}
2203 
2204 	if (adapter->ioport != NULL) {
2205 		bus_release_resource(dev, SYS_RES_IOPORT,
2206 		    rman_get_rid(adapter->ioport), adapter->ioport);
2207 		adapter->ioport = NULL;
2208 	}
2209 }
2210 
2211 /* Set up MSI or MSI-X */
2212 static int
2213 em_setup_msix(if_ctx_t ctx)
2214 {
2215 	struct adapter *adapter = iflib_get_softc(ctx);
2216 
2217 	if (adapter->hw.mac.type == e1000_82574) {
2218 		em_enable_vectors_82574(ctx);
2219 	}
2220 	return (0);
2221 }
2222 
2223 /*********************************************************************
2224  *
2225  *  Workaround for SmartSpeed on 82541 and 82547 controllers
2226  *
2227  **********************************************************************/
2228 static void
2229 lem_smartspeed(struct adapter *adapter)
2230 {
2231 	u16 phy_tmp;
2232 
2233 	if (adapter->link_active || (adapter->hw.phy.type != e1000_phy_igp) ||
2234 	    adapter->hw.mac.autoneg == 0 ||
2235 	    (adapter->hw.phy.autoneg_advertised & ADVERTISE_1000_FULL) == 0)
2236 		return;
2237 
2238 	if (adapter->smartspeed == 0) {
2239 		/* If Master/Slave config fault is asserted twice,
2240 		 * we assume back-to-back */
2241 		e1000_read_phy_reg(&adapter->hw, PHY_1000T_STATUS, &phy_tmp);
2242 		if (!(phy_tmp & SR_1000T_MS_CONFIG_FAULT))
2243 			return;
2244 		e1000_read_phy_reg(&adapter->hw, PHY_1000T_STATUS, &phy_tmp);
2245 		if (phy_tmp & SR_1000T_MS_CONFIG_FAULT) {
2246 			e1000_read_phy_reg(&adapter->hw,
2247 			    PHY_1000T_CTRL, &phy_tmp);
2248 			if(phy_tmp & CR_1000T_MS_ENABLE) {
2249 				phy_tmp &= ~CR_1000T_MS_ENABLE;
2250 				e1000_write_phy_reg(&adapter->hw,
2251 				    PHY_1000T_CTRL, phy_tmp);
2252 				adapter->smartspeed++;
2253 				if(adapter->hw.mac.autoneg &&
2254 				   !e1000_copper_link_autoneg(&adapter->hw) &&
2255 				   !e1000_read_phy_reg(&adapter->hw,
2256 				    PHY_CONTROL, &phy_tmp)) {
2257 					phy_tmp |= (MII_CR_AUTO_NEG_EN |
2258 						    MII_CR_RESTART_AUTO_NEG);
2259 					e1000_write_phy_reg(&adapter->hw,
2260 					    PHY_CONTROL, phy_tmp);
2261 				}
2262 			}
2263 		}
2264 		return;
2265 	} else if(adapter->smartspeed == EM_SMARTSPEED_DOWNSHIFT) {
2266 		/* If still no link, perhaps using 2/3 pair cable */
2267 		e1000_read_phy_reg(&adapter->hw, PHY_1000T_CTRL, &phy_tmp);
2268 		phy_tmp |= CR_1000T_MS_ENABLE;
2269 		e1000_write_phy_reg(&adapter->hw, PHY_1000T_CTRL, phy_tmp);
2270 		if(adapter->hw.mac.autoneg &&
2271 		   !e1000_copper_link_autoneg(&adapter->hw) &&
2272 		   !e1000_read_phy_reg(&adapter->hw, PHY_CONTROL, &phy_tmp)) {
2273 			phy_tmp |= (MII_CR_AUTO_NEG_EN |
2274 				    MII_CR_RESTART_AUTO_NEG);
2275 			e1000_write_phy_reg(&adapter->hw, PHY_CONTROL, phy_tmp);
2276 		}
2277 	}
2278 	/* Restart process after EM_SMARTSPEED_MAX iterations */
2279 	if(adapter->smartspeed++ == EM_SMARTSPEED_MAX)
2280 		adapter->smartspeed = 0;
2281 }
2282 
2283 /*********************************************************************
2284  *
2285  *  Initialize the DMA Coalescing feature
2286  *
2287  **********************************************************************/
2288 static void
2289 igb_init_dmac(struct adapter *adapter, u32 pba)
2290 {
2291 	device_t	dev = adapter->dev;
2292 	struct e1000_hw *hw = &adapter->hw;
2293 	u32 		dmac, reg = ~E1000_DMACR_DMAC_EN;
2294 	u16		hwm;
2295 	u16		max_frame_size;
2296 
2297 	if (hw->mac.type == e1000_i211)
2298 		return;
2299 
2300 	max_frame_size = adapter->shared->isc_max_frame_size;
2301 	if (hw->mac.type > e1000_82580) {
2302 
2303 		if (adapter->dmac == 0) { /* Disabling it */
2304 			E1000_WRITE_REG(hw, E1000_DMACR, reg);
2305 			return;
2306 		} else
2307 			device_printf(dev, "DMA Coalescing enabled\n");
2308 
2309 		/* Set starting threshold */
2310 		E1000_WRITE_REG(hw, E1000_DMCTXTH, 0);
2311 
2312 		hwm = 64 * pba - max_frame_size / 16;
2313 		if (hwm < 64 * (pba - 6))
2314 			hwm = 64 * (pba - 6);
2315 		reg = E1000_READ_REG(hw, E1000_FCRTC);
2316 		reg &= ~E1000_FCRTC_RTH_COAL_MASK;
2317 		reg |= ((hwm << E1000_FCRTC_RTH_COAL_SHIFT)
2318 		    & E1000_FCRTC_RTH_COAL_MASK);
2319 		E1000_WRITE_REG(hw, E1000_FCRTC, reg);
2320 
2321 
2322 		dmac = pba - max_frame_size / 512;
2323 		if (dmac < pba - 10)
2324 			dmac = pba - 10;
2325 		reg = E1000_READ_REG(hw, E1000_DMACR);
2326 		reg &= ~E1000_DMACR_DMACTHR_MASK;
2327 		reg |= ((dmac << E1000_DMACR_DMACTHR_SHIFT)
2328 		    & E1000_DMACR_DMACTHR_MASK);
2329 
2330 		/* transition to L0x or L1 if available..*/
2331 		reg |= (E1000_DMACR_DMAC_EN | E1000_DMACR_DMAC_LX_MASK);
2332 
2333 		/* Check if status is 2.5Gb backplane connection
2334 		* before configuration of watchdog timer, which is
2335 		* in msec values in 12.8usec intervals
2336 		* watchdog timer= msec values in 32usec intervals
2337 		* for non 2.5Gb connection
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 |= ((adapter->dmac * 5) >> 6);
2344 			else
2345 				reg |= (adapter->dmac >> 5);
2346 		} else {
2347 			reg |= (adapter->dmac >> 5);
2348 		}
2349 
2350 		E1000_WRITE_REG(hw, E1000_DMACR, reg);
2351 
2352 		E1000_WRITE_REG(hw, E1000_DMCRTRH, 0);
2353 
2354 		/* Set the interval before transition */
2355 		reg = E1000_READ_REG(hw, E1000_DMCTLX);
2356 		if (hw->mac.type == e1000_i350)
2357 			reg |= IGB_DMCTLX_DCFLUSH_DIS;
2358 		/*
2359 		** in 2.5Gb connection, TTLX unit is 0.4 usec
2360 		** which is 0x4*2 = 0xA. But delay is still 4 usec
2361 		*/
2362 		if (hw->mac.type == e1000_i354) {
2363 			int status = E1000_READ_REG(hw, E1000_STATUS);
2364 			if ((status & E1000_STATUS_2P5_SKU) &&
2365 			    (!(status & E1000_STATUS_2P5_SKU_OVER)))
2366 				reg |= 0xA;
2367 			else
2368 				reg |= 0x4;
2369 		} else {
2370 			reg |= 0x4;
2371 		}
2372 
2373 		E1000_WRITE_REG(hw, E1000_DMCTLX, reg);
2374 
2375 		/* free space in tx packet buffer to wake from DMA coal */
2376 		E1000_WRITE_REG(hw, E1000_DMCTXTH, (IGB_TXPBSIZE -
2377 		    (2 * max_frame_size)) >> 6);
2378 
2379 		/* make low power state decision controlled by DMA coal */
2380 		reg = E1000_READ_REG(hw, E1000_PCIEMISC);
2381 		reg &= ~E1000_PCIEMISC_LX_DECISION;
2382 		E1000_WRITE_REG(hw, E1000_PCIEMISC, reg);
2383 
2384 	} else if (hw->mac.type == e1000_82580) {
2385 		u32 reg = E1000_READ_REG(hw, E1000_PCIEMISC);
2386 		E1000_WRITE_REG(hw, E1000_PCIEMISC,
2387 		    reg & ~E1000_PCIEMISC_LX_DECISION);
2388 		E1000_WRITE_REG(hw, E1000_DMACR, 0);
2389 	}
2390 }
2391 
2392 /*********************************************************************
2393  *
2394  *  Initialize the hardware to a configuration as specified by the
2395  *  adapter structure.
2396  *
2397  **********************************************************************/
2398 static void
2399 em_reset(if_ctx_t ctx)
2400 {
2401 	device_t dev = iflib_get_dev(ctx);
2402 	struct adapter *adapter = iflib_get_softc(ctx);
2403 	struct ifnet *ifp = iflib_get_ifp(ctx);
2404 	struct e1000_hw *hw = &adapter->hw;
2405 	u16 rx_buffer_size;
2406 	u32 pba;
2407 
2408 	INIT_DEBUGOUT("em_reset: begin");
2409 	/* Let the firmware know the OS is in control */
2410 	em_get_hw_control(adapter);
2411 
2412 	/* Set up smart power down as default off on newer adapters. */
2413 	if (!em_smart_pwr_down && (hw->mac.type == e1000_82571 ||
2414 	    hw->mac.type == e1000_82572)) {
2415 		u16 phy_tmp = 0;
2416 
2417 		/* Speed up time to link by disabling smart power down. */
2418 		e1000_read_phy_reg(hw, IGP02E1000_PHY_POWER_MGMT, &phy_tmp);
2419 		phy_tmp &= ~IGP02E1000_PM_SPD;
2420 		e1000_write_phy_reg(hw, IGP02E1000_PHY_POWER_MGMT, phy_tmp);
2421 	}
2422 
2423 	/*
2424 	 * Packet Buffer Allocation (PBA)
2425 	 * Writing PBA sets the receive portion of the buffer
2426 	 * the remainder is used for the transmit buffer.
2427 	 */
2428 	switch (hw->mac.type) {
2429 	/* Total Packet Buffer on these is 48K */
2430 	case e1000_82571:
2431 	case e1000_82572:
2432 	case e1000_80003es2lan:
2433 			pba = E1000_PBA_32K; /* 32K for Rx, 16K for Tx */
2434 		break;
2435 	case e1000_82573: /* 82573: Total Packet Buffer is 32K */
2436 			pba = E1000_PBA_12K; /* 12K for Rx, 20K for Tx */
2437 		break;
2438 	case e1000_82574:
2439 	case e1000_82583:
2440 			pba = E1000_PBA_20K; /* 20K for Rx, 20K for Tx */
2441 		break;
2442 	case e1000_ich8lan:
2443 		pba = E1000_PBA_8K;
2444 		break;
2445 	case e1000_ich9lan:
2446 	case e1000_ich10lan:
2447 		/* Boost Receive side for jumbo frames */
2448 		if (adapter->hw.mac.max_frame_size > 4096)
2449 			pba = E1000_PBA_14K;
2450 		else
2451 			pba = E1000_PBA_10K;
2452 		break;
2453 	case e1000_pchlan:
2454 	case e1000_pch2lan:
2455 	case e1000_pch_lpt:
2456 	case e1000_pch_spt:
2457 	case e1000_pch_cnp:
2458 		pba = E1000_PBA_26K;
2459 		break;
2460 	case e1000_82575:
2461 		pba = E1000_PBA_32K;
2462 		break;
2463 	case e1000_82576:
2464 	case e1000_vfadapt:
2465 		pba = E1000_READ_REG(hw, E1000_RXPBS);
2466 		pba &= E1000_RXPBS_SIZE_MASK_82576;
2467 		break;
2468 	case e1000_82580:
2469 	case e1000_i350:
2470 	case e1000_i354:
2471 	case e1000_vfadapt_i350:
2472 		pba = E1000_READ_REG(hw, E1000_RXPBS);
2473 		pba = e1000_rxpbs_adjust_82580(pba);
2474 		break;
2475 	case e1000_i210:
2476 	case e1000_i211:
2477 		pba = E1000_PBA_34K;
2478 		break;
2479 	default:
2480 		if (adapter->hw.mac.max_frame_size > 8192)
2481 			pba = E1000_PBA_40K; /* 40K for Rx, 24K for Tx */
2482 		else
2483 			pba = E1000_PBA_48K; /* 48K for Rx, 16K for Tx */
2484 	}
2485 
2486 	/* Special needs in case of Jumbo frames */
2487 	if ((hw->mac.type == e1000_82575) && (ifp->if_mtu > ETHERMTU)) {
2488 		u32 tx_space, min_tx, min_rx;
2489 		pba = E1000_READ_REG(hw, E1000_PBA);
2490 		tx_space = pba >> 16;
2491 		pba &= 0xffff;
2492 		min_tx = (adapter->hw.mac.max_frame_size +
2493 		    sizeof(struct e1000_tx_desc) - ETHERNET_FCS_SIZE) * 2;
2494 		min_tx = roundup2(min_tx, 1024);
2495 		min_tx >>= 10;
2496 		min_rx = adapter->hw.mac.max_frame_size;
2497 		min_rx = roundup2(min_rx, 1024);
2498 		min_rx >>= 10;
2499 		if (tx_space < min_tx &&
2500 		    ((min_tx - tx_space) < pba)) {
2501 			pba = pba - (min_tx - tx_space);
2502 			/*
2503 			 * if short on rx space, rx wins
2504 			 * and must trump tx adjustment
2505 			 */
2506 			if (pba < min_rx)
2507 				pba = min_rx;
2508 		}
2509 		E1000_WRITE_REG(hw, E1000_PBA, pba);
2510 	}
2511 
2512 	if (hw->mac.type < igb_mac_min)
2513 		E1000_WRITE_REG(&adapter->hw, E1000_PBA, pba);
2514 
2515 	INIT_DEBUGOUT1("em_reset: pba=%dK",pba);
2516 
2517 	/*
2518 	 * These parameters control the automatic generation (Tx) and
2519 	 * response (Rx) to Ethernet PAUSE frames.
2520 	 * - High water mark should allow for at least two frames to be
2521 	 *   received after sending an XOFF.
2522 	 * - Low water mark works best when it is very near the high water mark.
2523 	 *   This allows the receiver to restart by sending XON when it has
2524 	 *   drained a bit. Here we use an arbitrary value of 1500 which will
2525 	 *   restart after one full frame is pulled from the buffer. There
2526 	 *   could be several smaller frames in the buffer and if so they will
2527 	 *   not trigger the XON until their total number reduces the buffer
2528 	 *   by 1500.
2529 	 * - The pause time is fairly large at 1000 x 512ns = 512 usec.
2530 	 */
2531 	rx_buffer_size = (pba & 0xffff) << 10;
2532 	hw->fc.high_water = rx_buffer_size -
2533 	    roundup2(adapter->hw.mac.max_frame_size, 1024);
2534 	hw->fc.low_water = hw->fc.high_water - 1500;
2535 
2536 	if (adapter->fc) /* locally set flow control value? */
2537 		hw->fc.requested_mode = adapter->fc;
2538 	else
2539 		hw->fc.requested_mode = e1000_fc_full;
2540 
2541 	if (hw->mac.type == e1000_80003es2lan)
2542 		hw->fc.pause_time = 0xFFFF;
2543 	else
2544 		hw->fc.pause_time = EM_FC_PAUSE_TIME;
2545 
2546 	hw->fc.send_xon = TRUE;
2547 
2548 	/* Device specific overrides/settings */
2549 	switch (hw->mac.type) {
2550 	case e1000_pchlan:
2551 		/* Workaround: no TX flow ctrl for PCH */
2552 		hw->fc.requested_mode = e1000_fc_rx_pause;
2553 		hw->fc.pause_time = 0xFFFF; /* override */
2554 		if (if_getmtu(ifp) > ETHERMTU) {
2555 			hw->fc.high_water = 0x3500;
2556 			hw->fc.low_water = 0x1500;
2557 		} else {
2558 			hw->fc.high_water = 0x5000;
2559 			hw->fc.low_water = 0x3000;
2560 		}
2561 		hw->fc.refresh_time = 0x1000;
2562 		break;
2563 	case e1000_pch2lan:
2564 	case e1000_pch_lpt:
2565 	case e1000_pch_spt:
2566 	case e1000_pch_cnp:
2567 		hw->fc.high_water = 0x5C20;
2568 		hw->fc.low_water = 0x5048;
2569 		hw->fc.pause_time = 0x0650;
2570 		hw->fc.refresh_time = 0x0400;
2571 		/* Jumbos need adjusted PBA */
2572 		if (if_getmtu(ifp) > ETHERMTU)
2573 			E1000_WRITE_REG(hw, E1000_PBA, 12);
2574 		else
2575 			E1000_WRITE_REG(hw, E1000_PBA, 26);
2576 		break;
2577 	case e1000_82575:
2578 	case e1000_82576:
2579 		/* 8-byte granularity */
2580 		hw->fc.low_water = hw->fc.high_water - 8;
2581 		break;
2582 	case e1000_82580:
2583 	case e1000_i350:
2584 	case e1000_i354:
2585 	case e1000_i210:
2586 	case e1000_i211:
2587 	case e1000_vfadapt:
2588 	case e1000_vfadapt_i350:
2589 		/* 16-byte granularity */
2590 		hw->fc.low_water = hw->fc.high_water - 16;
2591 		break;
2592 	case e1000_ich9lan:
2593 	case e1000_ich10lan:
2594 		if (if_getmtu(ifp) > ETHERMTU) {
2595 			hw->fc.high_water = 0x2800;
2596 			hw->fc.low_water = hw->fc.high_water - 8;
2597 			break;
2598 		}
2599 		/* FALLTHROUGH */
2600 	default:
2601 		if (hw->mac.type == e1000_80003es2lan)
2602 			hw->fc.pause_time = 0xFFFF;
2603 		break;
2604 	}
2605 
2606 	/* Issue a global reset */
2607 	e1000_reset_hw(hw);
2608 	if (adapter->hw.mac.type >= igb_mac_min) {
2609 		E1000_WRITE_REG(hw, E1000_WUC, 0);
2610 	} else {
2611 		E1000_WRITE_REG(hw, E1000_WUFC, 0);
2612 		em_disable_aspm(adapter);
2613 	}
2614 	if (adapter->flags & IGB_MEDIA_RESET) {
2615 		e1000_setup_init_funcs(hw, TRUE);
2616 		e1000_get_bus_info(hw);
2617 		adapter->flags &= ~IGB_MEDIA_RESET;
2618 	}
2619 	/* and a re-init */
2620 	if (e1000_init_hw(hw) < 0) {
2621 		device_printf(dev, "Hardware Initialization Failed\n");
2622 		return;
2623 	}
2624 	if (adapter->hw.mac.type >= igb_mac_min)
2625 		igb_init_dmac(adapter, pba);
2626 
2627 	E1000_WRITE_REG(hw, E1000_VET, ETHERTYPE_VLAN);
2628 	e1000_get_phy_info(hw);
2629 	e1000_check_for_link(hw);
2630 }
2631 
2632 /*
2633  * Initialise the RSS mapping for NICs that support multiple transmit/
2634  * receive rings.
2635  */
2636 
2637 #define RSSKEYLEN 10
2638 static void
2639 em_initialize_rss_mapping(struct adapter *adapter)
2640 {
2641 	uint8_t  rss_key[4 * RSSKEYLEN];
2642 	uint32_t reta = 0;
2643 	struct e1000_hw	*hw = &adapter->hw;
2644 	int i;
2645 
2646 	/*
2647 	 * Configure RSS key
2648 	 */
2649 	arc4rand(rss_key, sizeof(rss_key), 0);
2650 	for (i = 0; i < RSSKEYLEN; ++i) {
2651 		uint32_t rssrk = 0;
2652 
2653 		rssrk = EM_RSSRK_VAL(rss_key, i);
2654 		E1000_WRITE_REG(hw,E1000_RSSRK(i), rssrk);
2655 	}
2656 
2657 	/*
2658 	 * Configure RSS redirect table in following fashion:
2659 	 * (hash & ring_cnt_mask) == rdr_table[(hash & rdr_table_mask)]
2660 	 */
2661 	for (i = 0; i < sizeof(reta); ++i) {
2662 		uint32_t q;
2663 
2664 		q = (i % adapter->rx_num_queues) << 7;
2665 		reta |= q << (8 * i);
2666 	}
2667 
2668 	for (i = 0; i < 32; ++i)
2669 		E1000_WRITE_REG(hw, E1000_RETA(i), reta);
2670 
2671 	E1000_WRITE_REG(hw, E1000_MRQC, E1000_MRQC_RSS_ENABLE_2Q |
2672 			E1000_MRQC_RSS_FIELD_IPV4_TCP |
2673 			E1000_MRQC_RSS_FIELD_IPV4 |
2674 			E1000_MRQC_RSS_FIELD_IPV6_TCP_EX |
2675 			E1000_MRQC_RSS_FIELD_IPV6_EX |
2676 			E1000_MRQC_RSS_FIELD_IPV6);
2677 }
2678 
2679 static void
2680 igb_initialize_rss_mapping(struct adapter *adapter)
2681 {
2682 	struct e1000_hw *hw = &adapter->hw;
2683 	int i;
2684 	int queue_id;
2685 	u32 reta;
2686 	u32 rss_key[10], mrqc, shift = 0;
2687 
2688 	/* XXX? */
2689 	if (adapter->hw.mac.type == e1000_82575)
2690 		shift = 6;
2691 
2692 	/*
2693 	 * The redirection table controls which destination
2694 	 * queue each bucket redirects traffic to.
2695 	 * Each DWORD represents four queues, with the LSB
2696 	 * being the first queue in the DWORD.
2697 	 *
2698 	 * This just allocates buckets to queues using round-robin
2699 	 * allocation.
2700 	 *
2701 	 * NOTE: It Just Happens to line up with the default
2702 	 * RSS allocation method.
2703 	 */
2704 
2705 	/* Warning FM follows */
2706 	reta = 0;
2707 	for (i = 0; i < 128; i++) {
2708 #ifdef RSS
2709 		queue_id = rss_get_indirection_to_bucket(i);
2710 		/*
2711 		 * If we have more queues than buckets, we'll
2712 		 * end up mapping buckets to a subset of the
2713 		 * queues.
2714 		 *
2715 		 * If we have more buckets than queues, we'll
2716 		 * end up instead assigning multiple buckets
2717 		 * to queues.
2718 		 *
2719 		 * Both are suboptimal, but we need to handle
2720 		 * the case so we don't go out of bounds
2721 		 * indexing arrays and such.
2722 		 */
2723 		queue_id = queue_id % adapter->rx_num_queues;
2724 #else
2725 		queue_id = (i % adapter->rx_num_queues);
2726 #endif
2727 		/* Adjust if required */
2728 		queue_id = queue_id << shift;
2729 
2730 		/*
2731 		 * The low 8 bits are for hash value (n+0);
2732 		 * The next 8 bits are for hash value (n+1), etc.
2733 		 */
2734 		reta = reta >> 8;
2735 		reta = reta | ( ((uint32_t) queue_id) << 24);
2736 		if ((i & 3) == 3) {
2737 			E1000_WRITE_REG(hw, E1000_RETA(i >> 2), reta);
2738 			reta = 0;
2739 		}
2740 	}
2741 
2742 	/* Now fill in hash table */
2743 
2744 	/*
2745 	 * MRQC: Multiple Receive Queues Command
2746 	 * Set queuing to RSS control, number depends on the device.
2747 	 */
2748 	mrqc = E1000_MRQC_ENABLE_RSS_8Q;
2749 
2750 #ifdef RSS
2751 	/* XXX ew typecasting */
2752 	rss_getkey((uint8_t *) &rss_key);
2753 #else
2754 	arc4rand(&rss_key, sizeof(rss_key), 0);
2755 #endif
2756 	for (i = 0; i < 10; i++)
2757 		E1000_WRITE_REG_ARRAY(hw, E1000_RSSRK(0), i, rss_key[i]);
2758 
2759 	/*
2760 	 * Configure the RSS fields to hash upon.
2761 	 */
2762 	mrqc |= (E1000_MRQC_RSS_FIELD_IPV4 |
2763 	    E1000_MRQC_RSS_FIELD_IPV4_TCP);
2764 	mrqc |= (E1000_MRQC_RSS_FIELD_IPV6 |
2765 	    E1000_MRQC_RSS_FIELD_IPV6_TCP);
2766 	mrqc |=( E1000_MRQC_RSS_FIELD_IPV4_UDP |
2767 	    E1000_MRQC_RSS_FIELD_IPV6_UDP);
2768 	mrqc |=( E1000_MRQC_RSS_FIELD_IPV6_UDP_EX |
2769 	    E1000_MRQC_RSS_FIELD_IPV6_TCP_EX);
2770 
2771 	E1000_WRITE_REG(hw, E1000_MRQC, mrqc);
2772 }
2773 
2774 /*********************************************************************
2775  *
2776  *  Setup networking device structure and register interface media.
2777  *
2778  **********************************************************************/
2779 static int
2780 em_setup_interface(if_ctx_t ctx)
2781 {
2782 	struct ifnet *ifp = iflib_get_ifp(ctx);
2783 	struct adapter *adapter = iflib_get_softc(ctx);
2784 	if_softc_ctx_t scctx = adapter->shared;
2785 
2786 	INIT_DEBUGOUT("em_setup_interface: begin");
2787 
2788 	/* Single Queue */
2789 	if (adapter->tx_num_queues == 1) {
2790 		if_setsendqlen(ifp, scctx->isc_ntxd[0] - 1);
2791 		if_setsendqready(ifp);
2792 	}
2793 
2794 	/*
2795 	 * Specify the media types supported by this adapter and register
2796 	 * callbacks to update media and link information
2797 	 */
2798 	if ((adapter->hw.phy.media_type == e1000_media_type_fiber) ||
2799 	    (adapter->hw.phy.media_type == e1000_media_type_internal_serdes)) {
2800 		u_char fiber_type = IFM_1000_SX;	/* default type */
2801 
2802 		if (adapter->hw.mac.type == e1000_82545)
2803 			fiber_type = IFM_1000_LX;
2804 		ifmedia_add(adapter->media, IFM_ETHER | fiber_type | IFM_FDX, 0, NULL);
2805 		ifmedia_add(adapter->media, IFM_ETHER | fiber_type, 0, NULL);
2806 	} else {
2807 		ifmedia_add(adapter->media, IFM_ETHER | IFM_10_T, 0, NULL);
2808 		ifmedia_add(adapter->media, IFM_ETHER | IFM_10_T | IFM_FDX, 0, NULL);
2809 		ifmedia_add(adapter->media, IFM_ETHER | IFM_100_TX, 0, NULL);
2810 		ifmedia_add(adapter->media, IFM_ETHER | IFM_100_TX | IFM_FDX, 0, NULL);
2811 		if (adapter->hw.phy.type != e1000_phy_ife) {
2812 			ifmedia_add(adapter->media, IFM_ETHER | IFM_1000_T | IFM_FDX, 0, NULL);
2813 			ifmedia_add(adapter->media, IFM_ETHER | IFM_1000_T, 0, NULL);
2814 		}
2815 	}
2816 	ifmedia_add(adapter->media, IFM_ETHER | IFM_AUTO, 0, NULL);
2817 	ifmedia_set(adapter->media, IFM_ETHER | IFM_AUTO);
2818 	return (0);
2819 }
2820 
2821 static int
2822 em_if_tx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int ntxqs, int ntxqsets)
2823 {
2824 	struct adapter *adapter = iflib_get_softc(ctx);
2825 	if_softc_ctx_t scctx = adapter->shared;
2826 	int error = E1000_SUCCESS;
2827 	struct em_tx_queue *que;
2828 	int i, j;
2829 
2830 	MPASS(adapter->tx_num_queues > 0);
2831 	MPASS(adapter->tx_num_queues == ntxqsets);
2832 
2833 	/* First allocate the top level queue structs */
2834 	if (!(adapter->tx_queues =
2835 	    (struct em_tx_queue *) malloc(sizeof(struct em_tx_queue) *
2836 	    adapter->tx_num_queues, M_DEVBUF, M_NOWAIT | M_ZERO))) {
2837 		device_printf(iflib_get_dev(ctx), "Unable to allocate queue memory\n");
2838 		return(ENOMEM);
2839 	}
2840 
2841 	for (i = 0, que = adapter->tx_queues; i < adapter->tx_num_queues; i++, que++) {
2842 		/* Set up some basics */
2843 
2844 		struct tx_ring *txr = &que->txr;
2845 		txr->adapter = que->adapter = adapter;
2846 		que->me = txr->me =  i;
2847 
2848 		/* Allocate report status array */
2849 		if (!(txr->tx_rsq = (qidx_t *) malloc(sizeof(qidx_t) * scctx->isc_ntxd[0], M_DEVBUF, M_NOWAIT | M_ZERO))) {
2850 			device_printf(iflib_get_dev(ctx), "failed to allocate rs_idxs memory\n");
2851 			error = ENOMEM;
2852 			goto fail;
2853 		}
2854 		for (j = 0; j < scctx->isc_ntxd[0]; j++)
2855 			txr->tx_rsq[j] = QIDX_INVALID;
2856 		/* get the virtual and physical address of the hardware queues */
2857 		txr->tx_base = (struct e1000_tx_desc *)vaddrs[i*ntxqs];
2858 		txr->tx_paddr = paddrs[i*ntxqs];
2859 	}
2860 
2861 	if (bootverbose)
2862 		device_printf(iflib_get_dev(ctx),
2863 		    "allocated for %d tx_queues\n", adapter->tx_num_queues);
2864 	return (0);
2865 fail:
2866 	em_if_queues_free(ctx);
2867 	return (error);
2868 }
2869 
2870 static int
2871 em_if_rx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int nrxqs, int nrxqsets)
2872 {
2873 	struct adapter *adapter = iflib_get_softc(ctx);
2874 	int error = E1000_SUCCESS;
2875 	struct em_rx_queue *que;
2876 	int i;
2877 
2878 	MPASS(adapter->rx_num_queues > 0);
2879 	MPASS(adapter->rx_num_queues == nrxqsets);
2880 
2881 	/* First allocate the top level queue structs */
2882 	if (!(adapter->rx_queues =
2883 	    (struct em_rx_queue *) malloc(sizeof(struct em_rx_queue) *
2884 	    adapter->rx_num_queues, M_DEVBUF, M_NOWAIT | M_ZERO))) {
2885 		device_printf(iflib_get_dev(ctx), "Unable to allocate queue memory\n");
2886 		error = ENOMEM;
2887 		goto fail;
2888 	}
2889 
2890 	for (i = 0, que = adapter->rx_queues; i < nrxqsets; i++, que++) {
2891 		/* Set up some basics */
2892 		struct rx_ring *rxr = &que->rxr;
2893 		rxr->adapter = que->adapter = adapter;
2894 		rxr->que = que;
2895 		que->me = rxr->me =  i;
2896 
2897 		/* get the virtual and physical address of the hardware queues */
2898 		rxr->rx_base = (union e1000_rx_desc_extended *)vaddrs[i*nrxqs];
2899 		rxr->rx_paddr = paddrs[i*nrxqs];
2900 	}
2901 
2902 	if (bootverbose)
2903 		device_printf(iflib_get_dev(ctx),
2904 		    "allocated for %d rx_queues\n", adapter->rx_num_queues);
2905 
2906 	return (0);
2907 fail:
2908 	em_if_queues_free(ctx);
2909 	return (error);
2910 }
2911 
2912 static void
2913 em_if_queues_free(if_ctx_t ctx)
2914 {
2915 	struct adapter *adapter = iflib_get_softc(ctx);
2916 	struct em_tx_queue *tx_que = adapter->tx_queues;
2917 	struct em_rx_queue *rx_que = adapter->rx_queues;
2918 
2919 	if (tx_que != NULL) {
2920 		for (int i = 0; i < adapter->tx_num_queues; i++, tx_que++) {
2921 			struct tx_ring *txr = &tx_que->txr;
2922 			if (txr->tx_rsq == NULL)
2923 				break;
2924 
2925 			free(txr->tx_rsq, M_DEVBUF);
2926 			txr->tx_rsq = NULL;
2927 		}
2928 		free(adapter->tx_queues, M_DEVBUF);
2929 		adapter->tx_queues = NULL;
2930 	}
2931 
2932 	if (rx_que != NULL) {
2933 		free(adapter->rx_queues, M_DEVBUF);
2934 		adapter->rx_queues = NULL;
2935 	}
2936 
2937 	em_release_hw_control(adapter);
2938 
2939 	if (adapter->mta != NULL) {
2940 		free(adapter->mta, M_DEVBUF);
2941 	}
2942 }
2943 
2944 /*********************************************************************
2945  *
2946  *  Enable transmit unit.
2947  *
2948  **********************************************************************/
2949 static void
2950 em_initialize_transmit_unit(if_ctx_t ctx)
2951 {
2952 	struct adapter *adapter = iflib_get_softc(ctx);
2953 	if_softc_ctx_t scctx = adapter->shared;
2954 	struct em_tx_queue *que;
2955 	struct tx_ring	*txr;
2956 	struct e1000_hw	*hw = &adapter->hw;
2957 	u32 tctl, txdctl = 0, tarc, tipg = 0;
2958 
2959 	INIT_DEBUGOUT("em_initialize_transmit_unit: begin");
2960 
2961 	for (int i = 0; i < adapter->tx_num_queues; i++, txr++) {
2962 		u64 bus_addr;
2963 		caddr_t offp, endp;
2964 
2965 		que = &adapter->tx_queues[i];
2966 		txr = &que->txr;
2967 		bus_addr = txr->tx_paddr;
2968 
2969 		/* Clear checksum offload context. */
2970 		offp = (caddr_t)&txr->csum_flags;
2971 		endp = (caddr_t)(txr + 1);
2972 		bzero(offp, endp - offp);
2973 
2974 		/* Base and Len of TX Ring */
2975 		E1000_WRITE_REG(hw, E1000_TDLEN(i),
2976 		    scctx->isc_ntxd[0] * sizeof(struct e1000_tx_desc));
2977 		E1000_WRITE_REG(hw, E1000_TDBAH(i),
2978 		    (u32)(bus_addr >> 32));
2979 		E1000_WRITE_REG(hw, E1000_TDBAL(i),
2980 		    (u32)bus_addr);
2981 		/* Init the HEAD/TAIL indices */
2982 		E1000_WRITE_REG(hw, E1000_TDT(i), 0);
2983 		E1000_WRITE_REG(hw, E1000_TDH(i), 0);
2984 
2985 		HW_DEBUGOUT2("Base = %x, Length = %x\n",
2986 		    E1000_READ_REG(&adapter->hw, E1000_TDBAL(i)),
2987 		    E1000_READ_REG(&adapter->hw, E1000_TDLEN(i)));
2988 
2989 		txdctl = 0; /* clear txdctl */
2990 		txdctl |= 0x1f; /* PTHRESH */
2991 		txdctl |= 1 << 8; /* HTHRESH */
2992 		txdctl |= 1 << 16;/* WTHRESH */
2993 		txdctl |= 1 << 22; /* Reserved bit 22 must always be 1 */
2994 		txdctl |= E1000_TXDCTL_GRAN;
2995 		txdctl |= 1 << 25; /* LWTHRESH */
2996 
2997 		E1000_WRITE_REG(hw, E1000_TXDCTL(i), txdctl);
2998 	}
2999 
3000 	/* Set the default values for the Tx Inter Packet Gap timer */
3001 	switch (adapter->hw.mac.type) {
3002 	case e1000_80003es2lan:
3003 		tipg = DEFAULT_82543_TIPG_IPGR1;
3004 		tipg |= DEFAULT_80003ES2LAN_TIPG_IPGR2 <<
3005 		    E1000_TIPG_IPGR2_SHIFT;
3006 		break;
3007 	case e1000_82542:
3008 		tipg = DEFAULT_82542_TIPG_IPGT;
3009 		tipg |= DEFAULT_82542_TIPG_IPGR1 << E1000_TIPG_IPGR1_SHIFT;
3010 		tipg |= DEFAULT_82542_TIPG_IPGR2 << E1000_TIPG_IPGR2_SHIFT;
3011 		break;
3012 	default:
3013 		if ((adapter->hw.phy.media_type == e1000_media_type_fiber) ||
3014 		    (adapter->hw.phy.media_type ==
3015 		    e1000_media_type_internal_serdes))
3016 			tipg = DEFAULT_82543_TIPG_IPGT_FIBER;
3017 		else
3018 			tipg = DEFAULT_82543_TIPG_IPGT_COPPER;
3019 		tipg |= DEFAULT_82543_TIPG_IPGR1 << E1000_TIPG_IPGR1_SHIFT;
3020 		tipg |= DEFAULT_82543_TIPG_IPGR2 << E1000_TIPG_IPGR2_SHIFT;
3021 	}
3022 
3023 	E1000_WRITE_REG(&adapter->hw, E1000_TIPG, tipg);
3024 	E1000_WRITE_REG(&adapter->hw, E1000_TIDV, adapter->tx_int_delay.value);
3025 
3026 	if(adapter->hw.mac.type >= e1000_82540)
3027 		E1000_WRITE_REG(&adapter->hw, E1000_TADV,
3028 		    adapter->tx_abs_int_delay.value);
3029 
3030 	if ((adapter->hw.mac.type == e1000_82571) ||
3031 	    (adapter->hw.mac.type == e1000_82572)) {
3032 		tarc = E1000_READ_REG(&adapter->hw, E1000_TARC(0));
3033 		tarc |= TARC_SPEED_MODE_BIT;
3034 		E1000_WRITE_REG(&adapter->hw, E1000_TARC(0), tarc);
3035 	} else if (adapter->hw.mac.type == e1000_80003es2lan) {
3036 		/* errata: program both queues to unweighted RR */
3037 		tarc = E1000_READ_REG(&adapter->hw, E1000_TARC(0));
3038 		tarc |= 1;
3039 		E1000_WRITE_REG(&adapter->hw, E1000_TARC(0), tarc);
3040 		tarc = E1000_READ_REG(&adapter->hw, E1000_TARC(1));
3041 		tarc |= 1;
3042 		E1000_WRITE_REG(&adapter->hw, E1000_TARC(1), tarc);
3043 	} else if (adapter->hw.mac.type == e1000_82574) {
3044 		tarc = E1000_READ_REG(&adapter->hw, E1000_TARC(0));
3045 		tarc |= TARC_ERRATA_BIT;
3046 		if ( adapter->tx_num_queues > 1) {
3047 			tarc |= (TARC_COMPENSATION_MODE | TARC_MQ_FIX);
3048 			E1000_WRITE_REG(&adapter->hw, E1000_TARC(0), tarc);
3049 			E1000_WRITE_REG(&adapter->hw, E1000_TARC(1), tarc);
3050 		} else
3051 			E1000_WRITE_REG(&adapter->hw, E1000_TARC(0), tarc);
3052 	}
3053 
3054 	if (adapter->tx_int_delay.value > 0)
3055 		adapter->txd_cmd |= E1000_TXD_CMD_IDE;
3056 
3057 	/* Program the Transmit Control Register */
3058 	tctl = E1000_READ_REG(&adapter->hw, E1000_TCTL);
3059 	tctl &= ~E1000_TCTL_CT;
3060 	tctl |= (E1000_TCTL_PSP | E1000_TCTL_RTLC | E1000_TCTL_EN |
3061 		   (E1000_COLLISION_THRESHOLD << E1000_CT_SHIFT));
3062 
3063 	if (adapter->hw.mac.type >= e1000_82571)
3064 		tctl |= E1000_TCTL_MULR;
3065 
3066 	/* This write will effectively turn on the transmit unit. */
3067 	E1000_WRITE_REG(&adapter->hw, E1000_TCTL, tctl);
3068 
3069 	/* SPT and KBL errata workarounds */
3070 	if (hw->mac.type == e1000_pch_spt) {
3071 		u32 reg;
3072 		reg = E1000_READ_REG(hw, E1000_IOSFPC);
3073 		reg |= E1000_RCTL_RDMTS_HEX;
3074 		E1000_WRITE_REG(hw, E1000_IOSFPC, reg);
3075 		/* i218-i219 Specification Update 1.5.4.5 */
3076 		reg = E1000_READ_REG(hw, E1000_TARC(0));
3077 		reg &= ~E1000_TARC0_CB_MULTIQ_3_REQ;
3078 		reg |= E1000_TARC0_CB_MULTIQ_2_REQ;
3079 		E1000_WRITE_REG(hw, E1000_TARC(0), reg);
3080 	}
3081 }
3082 
3083 /*********************************************************************
3084  *
3085  *  Enable receive unit.
3086  *
3087  **********************************************************************/
3088 
3089 static void
3090 em_initialize_receive_unit(if_ctx_t ctx)
3091 {
3092 	struct adapter *adapter = iflib_get_softc(ctx);
3093 	if_softc_ctx_t scctx = adapter->shared;
3094 	struct ifnet *ifp = iflib_get_ifp(ctx);
3095 	struct e1000_hw	*hw = &adapter->hw;
3096 	struct em_rx_queue *que;
3097 	int i;
3098 	u32 rctl, rxcsum, rfctl;
3099 
3100 	INIT_DEBUGOUT("em_initialize_receive_units: begin");
3101 
3102 	/*
3103 	 * Make sure receives are disabled while setting
3104 	 * up the descriptor ring
3105 	 */
3106 	rctl = E1000_READ_REG(hw, E1000_RCTL);
3107 	/* Do not disable if ever enabled on this hardware */
3108 	if ((hw->mac.type != e1000_82574) && (hw->mac.type != e1000_82583))
3109 		E1000_WRITE_REG(hw, E1000_RCTL, rctl & ~E1000_RCTL_EN);
3110 
3111 	/* Setup the Receive Control Register */
3112 	rctl &= ~(3 << E1000_RCTL_MO_SHIFT);
3113 	rctl |= E1000_RCTL_EN | E1000_RCTL_BAM |
3114 	    E1000_RCTL_LBM_NO | E1000_RCTL_RDMTS_HALF |
3115 	    (hw->mac.mc_filter_type << E1000_RCTL_MO_SHIFT);
3116 
3117 	/* Do not store bad packets */
3118 	rctl &= ~E1000_RCTL_SBP;
3119 
3120 	/* Enable Long Packet receive */
3121 	if (if_getmtu(ifp) > ETHERMTU)
3122 		rctl |= E1000_RCTL_LPE;
3123 	else
3124 		rctl &= ~E1000_RCTL_LPE;
3125 
3126 	/* Strip the CRC */
3127 	if (!em_disable_crc_stripping)
3128 		rctl |= E1000_RCTL_SECRC;
3129 
3130 	if (adapter->hw.mac.type >= e1000_82540) {
3131 		E1000_WRITE_REG(&adapter->hw, E1000_RADV,
3132 			    adapter->rx_abs_int_delay.value);
3133 
3134 		/*
3135 		 * Set the interrupt throttling rate. Value is calculated
3136 		 * as DEFAULT_ITR = 1/(MAX_INTS_PER_SEC * 256ns)
3137 		 */
3138 		E1000_WRITE_REG(hw, E1000_ITR, DEFAULT_ITR);
3139 	}
3140 	E1000_WRITE_REG(&adapter->hw, E1000_RDTR,
3141 	    adapter->rx_int_delay.value);
3142 
3143 	/* Use extended rx descriptor formats */
3144 	rfctl = E1000_READ_REG(hw, E1000_RFCTL);
3145 	rfctl |= E1000_RFCTL_EXTEN;
3146 	/*
3147 	 * When using MSI-X interrupts we need to throttle
3148 	 * using the EITR register (82574 only)
3149 	 */
3150 	if (hw->mac.type == e1000_82574) {
3151 		for (int i = 0; i < 4; i++)
3152 			E1000_WRITE_REG(hw, E1000_EITR_82574(i),
3153 			    DEFAULT_ITR);
3154 		/* Disable accelerated acknowledge */
3155 		rfctl |= E1000_RFCTL_ACK_DIS;
3156 	}
3157 	E1000_WRITE_REG(hw, E1000_RFCTL, rfctl);
3158 
3159 	rxcsum = E1000_READ_REG(hw, E1000_RXCSUM);
3160 	if (if_getcapenable(ifp) & IFCAP_RXCSUM &&
3161 	    adapter->hw.mac.type >= e1000_82543) {
3162 		if (adapter->tx_num_queues > 1) {
3163 			if (adapter->hw.mac.type >= igb_mac_min) {
3164 				rxcsum |= E1000_RXCSUM_PCSD;
3165 				if (hw->mac.type != e1000_82575)
3166 					rxcsum |= E1000_RXCSUM_CRCOFL;
3167 			} else
3168 				rxcsum |= E1000_RXCSUM_TUOFL |
3169 					E1000_RXCSUM_IPOFL |
3170 					E1000_RXCSUM_PCSD;
3171 		} else {
3172 			if (adapter->hw.mac.type >= igb_mac_min)
3173 				rxcsum |= E1000_RXCSUM_IPPCSE;
3174 			else
3175 				rxcsum |= E1000_RXCSUM_TUOFL | E1000_RXCSUM_IPOFL;
3176 			if (adapter->hw.mac.type > e1000_82575)
3177 				rxcsum |= E1000_RXCSUM_CRCOFL;
3178 		}
3179 	} else
3180 		rxcsum &= ~E1000_RXCSUM_TUOFL;
3181 
3182 	E1000_WRITE_REG(hw, E1000_RXCSUM, rxcsum);
3183 
3184 	if (adapter->rx_num_queues > 1) {
3185 		if (adapter->hw.mac.type >= igb_mac_min)
3186 			igb_initialize_rss_mapping(adapter);
3187 		else
3188 			em_initialize_rss_mapping(adapter);
3189 	}
3190 
3191 	/*
3192 	 * XXX TEMPORARY WORKAROUND: on some systems with 82573
3193 	 * long latencies are observed, like Lenovo X60. This
3194 	 * change eliminates the problem, but since having positive
3195 	 * values in RDTR is a known source of problems on other
3196 	 * platforms another solution is being sought.
3197 	 */
3198 	if (hw->mac.type == e1000_82573)
3199 		E1000_WRITE_REG(hw, E1000_RDTR, 0x20);
3200 
3201 	for (i = 0, que = adapter->rx_queues; i < adapter->rx_num_queues; i++, que++) {
3202 		struct rx_ring *rxr = &que->rxr;
3203 		/* Setup the Base and Length of the Rx Descriptor Ring */
3204 		u64 bus_addr = rxr->rx_paddr;
3205 #if 0
3206 		u32 rdt = adapter->rx_num_queues -1;  /* default */
3207 #endif
3208 
3209 		E1000_WRITE_REG(hw, E1000_RDLEN(i),
3210 		    scctx->isc_nrxd[0] * sizeof(union e1000_rx_desc_extended));
3211 		E1000_WRITE_REG(hw, E1000_RDBAH(i), (u32)(bus_addr >> 32));
3212 		E1000_WRITE_REG(hw, E1000_RDBAL(i), (u32)bus_addr);
3213 		/* Setup the Head and Tail Descriptor Pointers */
3214 		E1000_WRITE_REG(hw, E1000_RDH(i), 0);
3215 		E1000_WRITE_REG(hw, E1000_RDT(i), 0);
3216 	}
3217 
3218 	/*
3219 	 * Set PTHRESH for improved jumbo performance
3220 	 * According to 10.2.5.11 of Intel 82574 Datasheet,
3221 	 * RXDCTL(1) is written whenever RXDCTL(0) is written.
3222 	 * Only write to RXDCTL(1) if there is a need for different
3223 	 * settings.
3224 	 */
3225 
3226 	if (((adapter->hw.mac.type == e1000_ich9lan) ||
3227 	    (adapter->hw.mac.type == e1000_pch2lan) ||
3228 	    (adapter->hw.mac.type == e1000_ich10lan)) &&
3229 	    (if_getmtu(ifp) > ETHERMTU)) {
3230 		u32 rxdctl = E1000_READ_REG(hw, E1000_RXDCTL(0));
3231 		E1000_WRITE_REG(hw, E1000_RXDCTL(0), rxdctl | 3);
3232 	} else if (adapter->hw.mac.type == e1000_82574) {
3233 		for (int i = 0; i < adapter->rx_num_queues; i++) {
3234 			u32 rxdctl = E1000_READ_REG(hw, E1000_RXDCTL(i));
3235 			rxdctl |= 0x20; /* PTHRESH */
3236 			rxdctl |= 4 << 8; /* HTHRESH */
3237 			rxdctl |= 4 << 16;/* WTHRESH */
3238 			rxdctl |= 1 << 24; /* Switch to granularity */
3239 			E1000_WRITE_REG(hw, E1000_RXDCTL(i), rxdctl);
3240 		}
3241 	} else if (adapter->hw.mac.type >= igb_mac_min) {
3242 		u32 psize, srrctl = 0;
3243 
3244 		if (if_getmtu(ifp) > ETHERMTU) {
3245 			/* Set maximum packet len */
3246 			if (adapter->rx_mbuf_sz <= 4096) {
3247 				srrctl |= 4096 >> E1000_SRRCTL_BSIZEPKT_SHIFT;
3248 				rctl |= E1000_RCTL_SZ_4096 | E1000_RCTL_BSEX;
3249 			} else if (adapter->rx_mbuf_sz > 4096) {
3250 				srrctl |= 8192 >> E1000_SRRCTL_BSIZEPKT_SHIFT;
3251 				rctl |= E1000_RCTL_SZ_8192 | E1000_RCTL_BSEX;
3252 			}
3253 			psize = scctx->isc_max_frame_size;
3254 			/* are we on a vlan? */
3255 			if (ifp->if_vlantrunk != NULL)
3256 				psize += VLAN_TAG_SIZE;
3257 			E1000_WRITE_REG(&adapter->hw, E1000_RLPML, psize);
3258 		} else {
3259 			srrctl |= 2048 >> E1000_SRRCTL_BSIZEPKT_SHIFT;
3260 			rctl |= E1000_RCTL_SZ_2048;
3261 		}
3262 
3263 		/*
3264 		 * If TX flow control is disabled and there's >1 queue defined,
3265 		 * enable DROP.
3266 		 *
3267 		 * This drops frames rather than hanging the RX MAC for all queues.
3268 		 */
3269 		if ((adapter->rx_num_queues > 1) &&
3270 		    (adapter->fc == e1000_fc_none ||
3271 		     adapter->fc == e1000_fc_rx_pause)) {
3272 			srrctl |= E1000_SRRCTL_DROP_EN;
3273 		}
3274 			/* Setup the Base and Length of the Rx Descriptor Rings */
3275 		for (i = 0, que = adapter->rx_queues; i < adapter->rx_num_queues; i++, que++) {
3276 			struct rx_ring *rxr = &que->rxr;
3277 			u64 bus_addr = rxr->rx_paddr;
3278 			u32 rxdctl;
3279 
3280 #ifdef notyet
3281 			/* Configure for header split? -- ignore for now */
3282 			rxr->hdr_split = igb_header_split;
3283 #else
3284 			srrctl |= E1000_SRRCTL_DESCTYPE_ADV_ONEBUF;
3285 #endif
3286 
3287 			E1000_WRITE_REG(hw, E1000_RDLEN(i),
3288 					scctx->isc_nrxd[0] * sizeof(struct e1000_rx_desc));
3289 			E1000_WRITE_REG(hw, E1000_RDBAH(i),
3290 					(uint32_t)(bus_addr >> 32));
3291 			E1000_WRITE_REG(hw, E1000_RDBAL(i),
3292 					(uint32_t)bus_addr);
3293 			E1000_WRITE_REG(hw, E1000_SRRCTL(i), srrctl);
3294 			/* Enable this Queue */
3295 			rxdctl = E1000_READ_REG(hw, E1000_RXDCTL(i));
3296 			rxdctl |= E1000_RXDCTL_QUEUE_ENABLE;
3297 			rxdctl &= 0xFFF00000;
3298 			rxdctl |= IGB_RX_PTHRESH;
3299 			rxdctl |= IGB_RX_HTHRESH << 8;
3300 			rxdctl |= IGB_RX_WTHRESH << 16;
3301 			E1000_WRITE_REG(hw, E1000_RXDCTL(i), rxdctl);
3302 		}
3303 	} else if (adapter->hw.mac.type >= e1000_pch2lan) {
3304 		if (if_getmtu(ifp) > ETHERMTU)
3305 			e1000_lv_jumbo_workaround_ich8lan(hw, TRUE);
3306 		else
3307 			e1000_lv_jumbo_workaround_ich8lan(hw, FALSE);
3308 	}
3309 
3310 	/* Make sure VLAN Filters are off */
3311 	rctl &= ~E1000_RCTL_VFE;
3312 
3313 	if (adapter->hw.mac.type < igb_mac_min) {
3314 		if (adapter->rx_mbuf_sz == MCLBYTES)
3315 			rctl |= E1000_RCTL_SZ_2048;
3316 		else if (adapter->rx_mbuf_sz == MJUMPAGESIZE)
3317 			rctl |= E1000_RCTL_SZ_4096 | E1000_RCTL_BSEX;
3318 		else if (adapter->rx_mbuf_sz > MJUMPAGESIZE)
3319 			rctl |= E1000_RCTL_SZ_8192 | E1000_RCTL_BSEX;
3320 
3321 		/* ensure we clear use DTYPE of 00 here */
3322 		rctl &= ~0x00000C00;
3323 	}
3324 
3325 	/* Write out the settings */
3326 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
3327 
3328 	return;
3329 }
3330 
3331 static void
3332 em_if_vlan_register(if_ctx_t ctx, u16 vtag)
3333 {
3334 	struct adapter *adapter = iflib_get_softc(ctx);
3335 	u32 index, bit;
3336 
3337 	index = (vtag >> 5) & 0x7F;
3338 	bit = vtag & 0x1F;
3339 	adapter->shadow_vfta[index] |= (1 << bit);
3340 	++adapter->num_vlans;
3341 }
3342 
3343 static void
3344 em_if_vlan_unregister(if_ctx_t ctx, u16 vtag)
3345 {
3346 	struct adapter *adapter = iflib_get_softc(ctx);
3347 	u32 index, bit;
3348 
3349 	index = (vtag >> 5) & 0x7F;
3350 	bit = vtag & 0x1F;
3351 	adapter->shadow_vfta[index] &= ~(1 << bit);
3352 	--adapter->num_vlans;
3353 }
3354 
3355 static void
3356 em_setup_vlan_hw_support(struct adapter *adapter)
3357 {
3358 	struct e1000_hw *hw = &adapter->hw;
3359 	u32 reg;
3360 
3361 	/*
3362 	 * We get here thru init_locked, meaning
3363 	 * a soft reset, this has already cleared
3364 	 * the VFTA and other state, so if there
3365 	 * have been no vlan's registered do nothing.
3366 	 */
3367 	if (adapter->num_vlans == 0)
3368 		return;
3369 
3370 	/*
3371 	 * A soft reset zero's out the VFTA, so
3372 	 * we need to repopulate it now.
3373 	 */
3374 	for (int i = 0; i < EM_VFTA_SIZE; i++)
3375 		if (adapter->shadow_vfta[i] != 0)
3376 			E1000_WRITE_REG_ARRAY(hw, E1000_VFTA,
3377 			    i, adapter->shadow_vfta[i]);
3378 
3379 	reg = E1000_READ_REG(hw, E1000_CTRL);
3380 	reg |= E1000_CTRL_VME;
3381 	E1000_WRITE_REG(hw, E1000_CTRL, reg);
3382 
3383 	/* Enable the Filter Table */
3384 	reg = E1000_READ_REG(hw, E1000_RCTL);
3385 	reg &= ~E1000_RCTL_CFIEN;
3386 	reg |= E1000_RCTL_VFE;
3387 	E1000_WRITE_REG(hw, E1000_RCTL, reg);
3388 }
3389 
3390 static void
3391 em_if_enable_intr(if_ctx_t ctx)
3392 {
3393 	struct adapter *adapter = iflib_get_softc(ctx);
3394 	struct e1000_hw *hw = &adapter->hw;
3395 	u32 ims_mask = IMS_ENABLE_MASK;
3396 
3397 	if (hw->mac.type == e1000_82574) {
3398 		E1000_WRITE_REG(hw, EM_EIAC, EM_MSIX_MASK);
3399 		ims_mask |= adapter->ims;
3400 	} else if (adapter->intr_type == IFLIB_INTR_MSIX && hw->mac.type >= igb_mac_min)  {
3401 		u32 mask = (adapter->que_mask | adapter->link_mask);
3402 
3403 		E1000_WRITE_REG(&adapter->hw, E1000_EIAC, mask);
3404 		E1000_WRITE_REG(&adapter->hw, E1000_EIAM, mask);
3405 		E1000_WRITE_REG(&adapter->hw, E1000_EIMS, mask);
3406 		ims_mask = E1000_IMS_LSC;
3407 	}
3408 
3409 	E1000_WRITE_REG(hw, E1000_IMS, ims_mask);
3410 }
3411 
3412 static void
3413 em_if_disable_intr(if_ctx_t ctx)
3414 {
3415 	struct adapter *adapter = iflib_get_softc(ctx);
3416 	struct e1000_hw *hw = &adapter->hw;
3417 
3418 	if (adapter->intr_type == IFLIB_INTR_MSIX) {
3419 		if (hw->mac.type >= igb_mac_min)
3420 			E1000_WRITE_REG(&adapter->hw, E1000_EIMC, ~0);
3421 		E1000_WRITE_REG(&adapter->hw, E1000_EIAC, 0);
3422 	}
3423 	E1000_WRITE_REG(&adapter->hw, E1000_IMC, 0xffffffff);
3424 }
3425 
3426 /*
3427  * Bit of a misnomer, what this really means is
3428  * to enable OS management of the system... aka
3429  * to disable special hardware management features
3430  */
3431 static void
3432 em_init_manageability(struct adapter *adapter)
3433 {
3434 	/* A shared code workaround */
3435 #define E1000_82542_MANC2H E1000_MANC2H
3436 	if (adapter->has_manage) {
3437 		int manc2h = E1000_READ_REG(&adapter->hw, E1000_MANC2H);
3438 		int manc = E1000_READ_REG(&adapter->hw, E1000_MANC);
3439 
3440 		/* disable hardware interception of ARP */
3441 		manc &= ~(E1000_MANC_ARP_EN);
3442 
3443 		/* enable receiving management packets to the host */
3444 		manc |= E1000_MANC_EN_MNG2HOST;
3445 #define E1000_MNG2HOST_PORT_623 (1 << 5)
3446 #define E1000_MNG2HOST_PORT_664 (1 << 6)
3447 		manc2h |= E1000_MNG2HOST_PORT_623;
3448 		manc2h |= E1000_MNG2HOST_PORT_664;
3449 		E1000_WRITE_REG(&adapter->hw, E1000_MANC2H, manc2h);
3450 		E1000_WRITE_REG(&adapter->hw, E1000_MANC, manc);
3451 	}
3452 }
3453 
3454 /*
3455  * Give control back to hardware management
3456  * controller if there is one.
3457  */
3458 static void
3459 em_release_manageability(struct adapter *adapter)
3460 {
3461 	if (adapter->has_manage) {
3462 		int manc = E1000_READ_REG(&adapter->hw, E1000_MANC);
3463 
3464 		/* re-enable hardware interception of ARP */
3465 		manc |= E1000_MANC_ARP_EN;
3466 		manc &= ~E1000_MANC_EN_MNG2HOST;
3467 
3468 		E1000_WRITE_REG(&adapter->hw, E1000_MANC, manc);
3469 	}
3470 }
3471 
3472 /*
3473  * em_get_hw_control sets the {CTRL_EXT|FWSM}:DRV_LOAD bit.
3474  * For ASF and Pass Through versions of f/w this means
3475  * that the driver is loaded. For AMT version type f/w
3476  * this means that the network i/f is open.
3477  */
3478 static void
3479 em_get_hw_control(struct adapter *adapter)
3480 {
3481 	u32 ctrl_ext, swsm;
3482 
3483 	if (adapter->vf_ifp)
3484 		return;
3485 
3486 	if (adapter->hw.mac.type == e1000_82573) {
3487 		swsm = E1000_READ_REG(&adapter->hw, E1000_SWSM);
3488 		E1000_WRITE_REG(&adapter->hw, E1000_SWSM,
3489 		    swsm | E1000_SWSM_DRV_LOAD);
3490 		return;
3491 	}
3492 	/* else */
3493 	ctrl_ext = E1000_READ_REG(&adapter->hw, E1000_CTRL_EXT);
3494 	E1000_WRITE_REG(&adapter->hw, E1000_CTRL_EXT,
3495 	    ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
3496 }
3497 
3498 /*
3499  * em_release_hw_control resets {CTRL_EXT|FWSM}:DRV_LOAD bit.
3500  * For ASF and Pass Through versions of f/w this means that
3501  * the driver is no longer loaded. For AMT versions of the
3502  * f/w this means that the network i/f is closed.
3503  */
3504 static void
3505 em_release_hw_control(struct adapter *adapter)
3506 {
3507 	u32 ctrl_ext, swsm;
3508 
3509 	if (!adapter->has_manage)
3510 		return;
3511 
3512 	if (adapter->hw.mac.type == e1000_82573) {
3513 		swsm = E1000_READ_REG(&adapter->hw, E1000_SWSM);
3514 		E1000_WRITE_REG(&adapter->hw, E1000_SWSM,
3515 		    swsm & ~E1000_SWSM_DRV_LOAD);
3516 		return;
3517 	}
3518 	/* else */
3519 	ctrl_ext = E1000_READ_REG(&adapter->hw, E1000_CTRL_EXT);
3520 	E1000_WRITE_REG(&adapter->hw, E1000_CTRL_EXT,
3521 	    ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
3522 	return;
3523 }
3524 
3525 static int
3526 em_is_valid_ether_addr(u8 *addr)
3527 {
3528 	char zero_addr[6] = { 0, 0, 0, 0, 0, 0 };
3529 
3530 	if ((addr[0] & 1) || (!bcmp(addr, zero_addr, ETHER_ADDR_LEN))) {
3531 		return (FALSE);
3532 	}
3533 
3534 	return (TRUE);
3535 }
3536 
3537 /*
3538 ** Parse the interface capabilities with regard
3539 ** to both system management and wake-on-lan for
3540 ** later use.
3541 */
3542 static void
3543 em_get_wakeup(if_ctx_t ctx)
3544 {
3545 	struct adapter *adapter = iflib_get_softc(ctx);
3546 	device_t dev = iflib_get_dev(ctx);
3547 	u16 eeprom_data = 0, device_id, apme_mask;
3548 
3549 	adapter->has_manage = e1000_enable_mng_pass_thru(&adapter->hw);
3550 	apme_mask = EM_EEPROM_APME;
3551 
3552 	switch (adapter->hw.mac.type) {
3553 	case e1000_82542:
3554 	case e1000_82543:
3555 		break;
3556 	case e1000_82544:
3557 		e1000_read_nvm(&adapter->hw,
3558 		    NVM_INIT_CONTROL2_REG, 1, &eeprom_data);
3559 		apme_mask = EM_82544_APME;
3560 		break;
3561 	case e1000_82546:
3562 	case e1000_82546_rev_3:
3563 		if (adapter->hw.bus.func == 1) {
3564 			e1000_read_nvm(&adapter->hw,
3565 			    NVM_INIT_CONTROL3_PORT_B, 1, &eeprom_data);
3566 			break;
3567 		} else
3568 			e1000_read_nvm(&adapter->hw,
3569 			    NVM_INIT_CONTROL3_PORT_A, 1, &eeprom_data);
3570 		break;
3571 	case e1000_82573:
3572 	case e1000_82583:
3573 		adapter->has_amt = TRUE;
3574 		/* FALLTHROUGH */
3575 	case e1000_82571:
3576 	case e1000_82572:
3577 	case e1000_80003es2lan:
3578 		if (adapter->hw.bus.func == 1) {
3579 			e1000_read_nvm(&adapter->hw,
3580 			    NVM_INIT_CONTROL3_PORT_B, 1, &eeprom_data);
3581 			break;
3582 		} else
3583 			e1000_read_nvm(&adapter->hw,
3584 			    NVM_INIT_CONTROL3_PORT_A, 1, &eeprom_data);
3585 		break;
3586 	case e1000_ich8lan:
3587 	case e1000_ich9lan:
3588 	case e1000_ich10lan:
3589 	case e1000_pchlan:
3590 	case e1000_pch2lan:
3591 	case e1000_pch_lpt:
3592 	case e1000_pch_spt:
3593 	case e1000_82575:	/* listing all igb devices */
3594 	case e1000_82576:
3595 	case e1000_82580:
3596 	case e1000_i350:
3597 	case e1000_i354:
3598 	case e1000_i210:
3599 	case e1000_i211:
3600 	case e1000_vfadapt:
3601 	case e1000_vfadapt_i350:
3602 		apme_mask = E1000_WUC_APME;
3603 		adapter->has_amt = TRUE;
3604 		eeprom_data = E1000_READ_REG(&adapter->hw, E1000_WUC);
3605 		break;
3606 	default:
3607 		e1000_read_nvm(&adapter->hw,
3608 		    NVM_INIT_CONTROL3_PORT_A, 1, &eeprom_data);
3609 		break;
3610 	}
3611 	if (eeprom_data & apme_mask)
3612 		adapter->wol = (E1000_WUFC_MAG | E1000_WUFC_MC);
3613 	/*
3614 	 * We have the eeprom settings, now apply the special cases
3615 	 * where the eeprom may be wrong or the board won't support
3616 	 * wake on lan on a particular port
3617 	 */
3618 	device_id = pci_get_device(dev);
3619 	switch (device_id) {
3620 	case E1000_DEV_ID_82546GB_PCIE:
3621 		adapter->wol = 0;
3622 		break;
3623 	case E1000_DEV_ID_82546EB_FIBER:
3624 	case E1000_DEV_ID_82546GB_FIBER:
3625 		/* Wake events only supported on port A for dual fiber
3626 		 * regardless of eeprom setting */
3627 		if (E1000_READ_REG(&adapter->hw, E1000_STATUS) &
3628 		    E1000_STATUS_FUNC_1)
3629 			adapter->wol = 0;
3630 		break;
3631 	case E1000_DEV_ID_82546GB_QUAD_COPPER_KSP3:
3632 		/* if quad port adapter, disable WoL on all but port A */
3633 		if (global_quad_port_a != 0)
3634 			adapter->wol = 0;
3635 		/* Reset for multiple quad port adapters */
3636 		if (++global_quad_port_a == 4)
3637 			global_quad_port_a = 0;
3638 		break;
3639 	case E1000_DEV_ID_82571EB_FIBER:
3640 		/* Wake events only supported on port A for dual fiber
3641 		 * regardless of eeprom setting */
3642 		if (E1000_READ_REG(&adapter->hw, E1000_STATUS) &
3643 		    E1000_STATUS_FUNC_1)
3644 			adapter->wol = 0;
3645 		break;
3646 	case E1000_DEV_ID_82571EB_QUAD_COPPER:
3647 	case E1000_DEV_ID_82571EB_QUAD_FIBER:
3648 	case E1000_DEV_ID_82571EB_QUAD_COPPER_LP:
3649 		/* if quad port adapter, disable WoL on all but port A */
3650 		if (global_quad_port_a != 0)
3651 			adapter->wol = 0;
3652 		/* Reset for multiple quad port adapters */
3653 		if (++global_quad_port_a == 4)
3654 			global_quad_port_a = 0;
3655 		break;
3656 	}
3657 	return;
3658 }
3659 
3660 
3661 /*
3662  * Enable PCI Wake On Lan capability
3663  */
3664 static void
3665 em_enable_wakeup(if_ctx_t ctx)
3666 {
3667 	struct adapter *adapter = iflib_get_softc(ctx);
3668 	device_t dev = iflib_get_dev(ctx);
3669 	if_t ifp = iflib_get_ifp(ctx);
3670 	int error = 0;
3671 	u32 pmc, ctrl, ctrl_ext, rctl;
3672 	u16 status;
3673 
3674 	if (pci_find_cap(dev, PCIY_PMG, &pmc) != 0)
3675 		return;
3676 
3677 	/*
3678 	 * Determine type of Wakeup: note that wol
3679 	 * is set with all bits on by default.
3680 	 */
3681 	if ((if_getcapenable(ifp) & IFCAP_WOL_MAGIC) == 0)
3682 		adapter->wol &= ~E1000_WUFC_MAG;
3683 
3684 	if ((if_getcapenable(ifp) & IFCAP_WOL_UCAST) == 0)
3685 		adapter->wol &= ~E1000_WUFC_EX;
3686 
3687 	if ((if_getcapenable(ifp) & IFCAP_WOL_MCAST) == 0)
3688 		adapter->wol &= ~E1000_WUFC_MC;
3689 	else {
3690 		rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
3691 		rctl |= E1000_RCTL_MPE;
3692 		E1000_WRITE_REG(&adapter->hw, E1000_RCTL, rctl);
3693 	}
3694 
3695 	if (!(adapter->wol & (E1000_WUFC_EX | E1000_WUFC_MAG | E1000_WUFC_MC)))
3696 		goto pme;
3697 
3698 	/* Advertise the wakeup capability */
3699 	ctrl = E1000_READ_REG(&adapter->hw, E1000_CTRL);
3700 	ctrl |= (E1000_CTRL_SWDPIN2 | E1000_CTRL_SWDPIN3);
3701 	E1000_WRITE_REG(&adapter->hw, E1000_CTRL, ctrl);
3702 
3703 	/* Keep the laser running on Fiber adapters */
3704 	if (adapter->hw.phy.media_type == e1000_media_type_fiber ||
3705 	    adapter->hw.phy.media_type == e1000_media_type_internal_serdes) {
3706 		ctrl_ext = E1000_READ_REG(&adapter->hw, E1000_CTRL_EXT);
3707 		ctrl_ext |= E1000_CTRL_EXT_SDP3_DATA;
3708 		E1000_WRITE_REG(&adapter->hw, E1000_CTRL_EXT, ctrl_ext);
3709 	}
3710 
3711 	if ((adapter->hw.mac.type == e1000_ich8lan) ||
3712 	    (adapter->hw.mac.type == e1000_pchlan) ||
3713 	    (adapter->hw.mac.type == e1000_ich9lan) ||
3714 	    (adapter->hw.mac.type == e1000_ich10lan))
3715 		e1000_suspend_workarounds_ich8lan(&adapter->hw);
3716 
3717 	if ( adapter->hw.mac.type >= e1000_pchlan) {
3718 		error = em_enable_phy_wakeup(adapter);
3719 		if (error)
3720 			goto pme;
3721 	} else {
3722 		/* Enable wakeup by the MAC */
3723 		E1000_WRITE_REG(&adapter->hw, E1000_WUC, E1000_WUC_PME_EN);
3724 		E1000_WRITE_REG(&adapter->hw, E1000_WUFC, adapter->wol);
3725 	}
3726 
3727 	if (adapter->hw.phy.type == e1000_phy_igp_3)
3728 		e1000_igp3_phy_powerdown_workaround_ich8lan(&adapter->hw);
3729 
3730 pme:
3731 	status = pci_read_config(dev, pmc + PCIR_POWER_STATUS, 2);
3732 	status &= ~(PCIM_PSTAT_PME | PCIM_PSTAT_PMEENABLE);
3733 	if (!error && (if_getcapenable(ifp) & IFCAP_WOL))
3734 		status |= PCIM_PSTAT_PME | PCIM_PSTAT_PMEENABLE;
3735 	pci_write_config(dev, pmc + PCIR_POWER_STATUS, status, 2);
3736 
3737 	return;
3738 }
3739 
3740 /*
3741  * WOL in the newer chipset interfaces (pchlan)
3742  * require thing to be copied into the phy
3743  */
3744 static int
3745 em_enable_phy_wakeup(struct adapter *adapter)
3746 {
3747 	struct e1000_hw *hw = &adapter->hw;
3748 	u32 mreg, ret = 0;
3749 	u16 preg;
3750 
3751 	/* copy MAC RARs to PHY RARs */
3752 	e1000_copy_rx_addrs_to_phy_ich8lan(hw);
3753 
3754 	/* copy MAC MTA to PHY MTA */
3755 	for (int i = 0; i < adapter->hw.mac.mta_reg_count; i++) {
3756 		mreg = E1000_READ_REG_ARRAY(hw, E1000_MTA, i);
3757 		e1000_write_phy_reg(hw, BM_MTA(i), (u16)(mreg & 0xFFFF));
3758 		e1000_write_phy_reg(hw, BM_MTA(i) + 1,
3759 		    (u16)((mreg >> 16) & 0xFFFF));
3760 	}
3761 
3762 	/* configure PHY Rx Control register */
3763 	e1000_read_phy_reg(&adapter->hw, BM_RCTL, &preg);
3764 	mreg = E1000_READ_REG(hw, E1000_RCTL);
3765 	if (mreg & E1000_RCTL_UPE)
3766 		preg |= BM_RCTL_UPE;
3767 	if (mreg & E1000_RCTL_MPE)
3768 		preg |= BM_RCTL_MPE;
3769 	preg &= ~(BM_RCTL_MO_MASK);
3770 	if (mreg & E1000_RCTL_MO_3)
3771 		preg |= (((mreg & E1000_RCTL_MO_3) >> E1000_RCTL_MO_SHIFT)
3772 				<< BM_RCTL_MO_SHIFT);
3773 	if (mreg & E1000_RCTL_BAM)
3774 		preg |= BM_RCTL_BAM;
3775 	if (mreg & E1000_RCTL_PMCF)
3776 		preg |= BM_RCTL_PMCF;
3777 	mreg = E1000_READ_REG(hw, E1000_CTRL);
3778 	if (mreg & E1000_CTRL_RFCE)
3779 		preg |= BM_RCTL_RFCE;
3780 	e1000_write_phy_reg(&adapter->hw, BM_RCTL, preg);
3781 
3782 	/* enable PHY wakeup in MAC register */
3783 	E1000_WRITE_REG(hw, E1000_WUC,
3784 	    E1000_WUC_PHY_WAKE | E1000_WUC_PME_EN | E1000_WUC_APME);
3785 	E1000_WRITE_REG(hw, E1000_WUFC, adapter->wol);
3786 
3787 	/* configure and enable PHY wakeup in PHY registers */
3788 	e1000_write_phy_reg(&adapter->hw, BM_WUFC, adapter->wol);
3789 	e1000_write_phy_reg(&adapter->hw, BM_WUC, E1000_WUC_PME_EN);
3790 
3791 	/* activate PHY wakeup */
3792 	ret = hw->phy.ops.acquire(hw);
3793 	if (ret) {
3794 		printf("Could not acquire PHY\n");
3795 		return ret;
3796 	}
3797 	e1000_write_phy_reg_mdic(hw, IGP01E1000_PHY_PAGE_SELECT,
3798 	                         (BM_WUC_ENABLE_PAGE << IGP_PAGE_SHIFT));
3799 	ret = e1000_read_phy_reg_mdic(hw, BM_WUC_ENABLE_REG, &preg);
3800 	if (ret) {
3801 		printf("Could not read PHY page 769\n");
3802 		goto out;
3803 	}
3804 	preg |= BM_WUC_ENABLE_BIT | BM_WUC_HOST_WU_BIT;
3805 	ret = e1000_write_phy_reg_mdic(hw, BM_WUC_ENABLE_REG, preg);
3806 	if (ret)
3807 		printf("Could not set PHY Host Wakeup bit\n");
3808 out:
3809 	hw->phy.ops.release(hw);
3810 
3811 	return ret;
3812 }
3813 
3814 static void
3815 em_if_led_func(if_ctx_t ctx, int onoff)
3816 {
3817 	struct adapter *adapter = iflib_get_softc(ctx);
3818 
3819 	if (onoff) {
3820 		e1000_setup_led(&adapter->hw);
3821 		e1000_led_on(&adapter->hw);
3822 	} else {
3823 		e1000_led_off(&adapter->hw);
3824 		e1000_cleanup_led(&adapter->hw);
3825 	}
3826 }
3827 
3828 /*
3829  * Disable the L0S and L1 LINK states
3830  */
3831 static void
3832 em_disable_aspm(struct adapter *adapter)
3833 {
3834 	int base, reg;
3835 	u16 link_cap,link_ctrl;
3836 	device_t dev = adapter->dev;
3837 
3838 	switch (adapter->hw.mac.type) {
3839 	case e1000_82573:
3840 	case e1000_82574:
3841 	case e1000_82583:
3842 		break;
3843 	default:
3844 		return;
3845 	}
3846 	if (pci_find_cap(dev, PCIY_EXPRESS, &base) != 0)
3847 		return;
3848 	reg = base + PCIER_LINK_CAP;
3849 	link_cap = pci_read_config(dev, reg, 2);
3850 	if ((link_cap & PCIEM_LINK_CAP_ASPM) == 0)
3851 		return;
3852 	reg = base + PCIER_LINK_CTL;
3853 	link_ctrl = pci_read_config(dev, reg, 2);
3854 	link_ctrl &= ~PCIEM_LINK_CTL_ASPMC;
3855 	pci_write_config(dev, reg, link_ctrl, 2);
3856 	return;
3857 }
3858 
3859 /**********************************************************************
3860  *
3861  *  Update the board statistics counters.
3862  *
3863  **********************************************************************/
3864 static void
3865 em_update_stats_counters(struct adapter *adapter)
3866 {
3867 
3868 	if(adapter->hw.phy.media_type == e1000_media_type_copper ||
3869 	   (E1000_READ_REG(&adapter->hw, E1000_STATUS) & E1000_STATUS_LU)) {
3870 		adapter->stats.symerrs += E1000_READ_REG(&adapter->hw, E1000_SYMERRS);
3871 		adapter->stats.sec += E1000_READ_REG(&adapter->hw, E1000_SEC);
3872 	}
3873 	adapter->stats.crcerrs += E1000_READ_REG(&adapter->hw, E1000_CRCERRS);
3874 	adapter->stats.mpc += E1000_READ_REG(&adapter->hw, E1000_MPC);
3875 	adapter->stats.scc += E1000_READ_REG(&adapter->hw, E1000_SCC);
3876 	adapter->stats.ecol += E1000_READ_REG(&adapter->hw, E1000_ECOL);
3877 
3878 	adapter->stats.mcc += E1000_READ_REG(&adapter->hw, E1000_MCC);
3879 	adapter->stats.latecol += E1000_READ_REG(&adapter->hw, E1000_LATECOL);
3880 	adapter->stats.colc += E1000_READ_REG(&adapter->hw, E1000_COLC);
3881 	adapter->stats.dc += E1000_READ_REG(&adapter->hw, E1000_DC);
3882 	adapter->stats.rlec += E1000_READ_REG(&adapter->hw, E1000_RLEC);
3883 	adapter->stats.xonrxc += E1000_READ_REG(&adapter->hw, E1000_XONRXC);
3884 	adapter->stats.xontxc += E1000_READ_REG(&adapter->hw, E1000_XONTXC);
3885 	adapter->stats.xoffrxc += E1000_READ_REG(&adapter->hw, E1000_XOFFRXC);
3886 	/*
3887 	 ** For watchdog management we need to know if we have been
3888 	 ** paused during the last interval, so capture that here.
3889 	*/
3890 	adapter->shared->isc_pause_frames = adapter->stats.xoffrxc;
3891 	adapter->stats.xofftxc += E1000_READ_REG(&adapter->hw, E1000_XOFFTXC);
3892 	adapter->stats.fcruc += E1000_READ_REG(&adapter->hw, E1000_FCRUC);
3893 	adapter->stats.prc64 += E1000_READ_REG(&adapter->hw, E1000_PRC64);
3894 	adapter->stats.prc127 += E1000_READ_REG(&adapter->hw, E1000_PRC127);
3895 	adapter->stats.prc255 += E1000_READ_REG(&adapter->hw, E1000_PRC255);
3896 	adapter->stats.prc511 += E1000_READ_REG(&adapter->hw, E1000_PRC511);
3897 	adapter->stats.prc1023 += E1000_READ_REG(&adapter->hw, E1000_PRC1023);
3898 	adapter->stats.prc1522 += E1000_READ_REG(&adapter->hw, E1000_PRC1522);
3899 	adapter->stats.gprc += E1000_READ_REG(&adapter->hw, E1000_GPRC);
3900 	adapter->stats.bprc += E1000_READ_REG(&adapter->hw, E1000_BPRC);
3901 	adapter->stats.mprc += E1000_READ_REG(&adapter->hw, E1000_MPRC);
3902 	adapter->stats.gptc += E1000_READ_REG(&adapter->hw, E1000_GPTC);
3903 
3904 	/* For the 64-bit byte counters the low dword must be read first. */
3905 	/* Both registers clear on the read of the high dword */
3906 
3907 	adapter->stats.gorc += E1000_READ_REG(&adapter->hw, E1000_GORCL) +
3908 	    ((u64)E1000_READ_REG(&adapter->hw, E1000_GORCH) << 32);
3909 	adapter->stats.gotc += E1000_READ_REG(&adapter->hw, E1000_GOTCL) +
3910 	    ((u64)E1000_READ_REG(&adapter->hw, E1000_GOTCH) << 32);
3911 
3912 	adapter->stats.rnbc += E1000_READ_REG(&adapter->hw, E1000_RNBC);
3913 	adapter->stats.ruc += E1000_READ_REG(&adapter->hw, E1000_RUC);
3914 	adapter->stats.rfc += E1000_READ_REG(&adapter->hw, E1000_RFC);
3915 	adapter->stats.roc += E1000_READ_REG(&adapter->hw, E1000_ROC);
3916 	adapter->stats.rjc += E1000_READ_REG(&adapter->hw, E1000_RJC);
3917 
3918 	adapter->stats.tor += E1000_READ_REG(&adapter->hw, E1000_TORH);
3919 	adapter->stats.tot += E1000_READ_REG(&adapter->hw, E1000_TOTH);
3920 
3921 	adapter->stats.tpr += E1000_READ_REG(&adapter->hw, E1000_TPR);
3922 	adapter->stats.tpt += E1000_READ_REG(&adapter->hw, E1000_TPT);
3923 	adapter->stats.ptc64 += E1000_READ_REG(&adapter->hw, E1000_PTC64);
3924 	adapter->stats.ptc127 += E1000_READ_REG(&adapter->hw, E1000_PTC127);
3925 	adapter->stats.ptc255 += E1000_READ_REG(&adapter->hw, E1000_PTC255);
3926 	adapter->stats.ptc511 += E1000_READ_REG(&adapter->hw, E1000_PTC511);
3927 	adapter->stats.ptc1023 += E1000_READ_REG(&adapter->hw, E1000_PTC1023);
3928 	adapter->stats.ptc1522 += E1000_READ_REG(&adapter->hw, E1000_PTC1522);
3929 	adapter->stats.mptc += E1000_READ_REG(&adapter->hw, E1000_MPTC);
3930 	adapter->stats.bptc += E1000_READ_REG(&adapter->hw, E1000_BPTC);
3931 
3932 	/* Interrupt Counts */
3933 
3934 	adapter->stats.iac += E1000_READ_REG(&adapter->hw, E1000_IAC);
3935 	adapter->stats.icrxptc += E1000_READ_REG(&adapter->hw, E1000_ICRXPTC);
3936 	adapter->stats.icrxatc += E1000_READ_REG(&adapter->hw, E1000_ICRXATC);
3937 	adapter->stats.ictxptc += E1000_READ_REG(&adapter->hw, E1000_ICTXPTC);
3938 	adapter->stats.ictxatc += E1000_READ_REG(&adapter->hw, E1000_ICTXATC);
3939 	adapter->stats.ictxqec += E1000_READ_REG(&adapter->hw, E1000_ICTXQEC);
3940 	adapter->stats.ictxqmtc += E1000_READ_REG(&adapter->hw, E1000_ICTXQMTC);
3941 	adapter->stats.icrxdmtc += E1000_READ_REG(&adapter->hw, E1000_ICRXDMTC);
3942 	adapter->stats.icrxoc += E1000_READ_REG(&adapter->hw, E1000_ICRXOC);
3943 
3944 	if (adapter->hw.mac.type >= e1000_82543) {
3945 		adapter->stats.algnerrc +=
3946 		E1000_READ_REG(&adapter->hw, E1000_ALGNERRC);
3947 		adapter->stats.rxerrc +=
3948 		E1000_READ_REG(&adapter->hw, E1000_RXERRC);
3949 		adapter->stats.tncrs +=
3950 		E1000_READ_REG(&adapter->hw, E1000_TNCRS);
3951 		adapter->stats.cexterr +=
3952 		E1000_READ_REG(&adapter->hw, E1000_CEXTERR);
3953 		adapter->stats.tsctc +=
3954 		E1000_READ_REG(&adapter->hw, E1000_TSCTC);
3955 		adapter->stats.tsctfc +=
3956 		E1000_READ_REG(&adapter->hw, E1000_TSCTFC);
3957 	}
3958 }
3959 
3960 static uint64_t
3961 em_if_get_counter(if_ctx_t ctx, ift_counter cnt)
3962 {
3963 	struct adapter *adapter = iflib_get_softc(ctx);
3964 	struct ifnet *ifp = iflib_get_ifp(ctx);
3965 
3966 	switch (cnt) {
3967 	case IFCOUNTER_COLLISIONS:
3968 		return (adapter->stats.colc);
3969 	case IFCOUNTER_IERRORS:
3970 		return (adapter->dropped_pkts + adapter->stats.rxerrc +
3971 		    adapter->stats.crcerrs + adapter->stats.algnerrc +
3972 		    adapter->stats.ruc + adapter->stats.roc +
3973 		    adapter->stats.mpc + adapter->stats.cexterr);
3974 	case IFCOUNTER_OERRORS:
3975 		return (adapter->stats.ecol + adapter->stats.latecol +
3976 		    adapter->watchdog_events);
3977 	default:
3978 		return (if_get_counter_default(ifp, cnt));
3979 	}
3980 }
3981 
3982 /* Export a single 32-bit register via a read-only sysctl. */
3983 static int
3984 em_sysctl_reg_handler(SYSCTL_HANDLER_ARGS)
3985 {
3986 	struct adapter *adapter;
3987 	u_int val;
3988 
3989 	adapter = oidp->oid_arg1;
3990 	val = E1000_READ_REG(&adapter->hw, oidp->oid_arg2);
3991 	return (sysctl_handle_int(oidp, &val, 0, req));
3992 }
3993 
3994 /*
3995  * Add sysctl variables, one per statistic, to the system.
3996  */
3997 static void
3998 em_add_hw_stats(struct adapter *adapter)
3999 {
4000 	device_t dev = iflib_get_dev(adapter->ctx);
4001 	struct em_tx_queue *tx_que = adapter->tx_queues;
4002 	struct em_rx_queue *rx_que = adapter->rx_queues;
4003 
4004 	struct sysctl_ctx_list *ctx = device_get_sysctl_ctx(dev);
4005 	struct sysctl_oid *tree = device_get_sysctl_tree(dev);
4006 	struct sysctl_oid_list *child = SYSCTL_CHILDREN(tree);
4007 	struct e1000_hw_stats *stats = &adapter->stats;
4008 
4009 	struct sysctl_oid *stat_node, *queue_node, *int_node;
4010 	struct sysctl_oid_list *stat_list, *queue_list, *int_list;
4011 
4012 #define QUEUE_NAME_LEN 32
4013 	char namebuf[QUEUE_NAME_LEN];
4014 
4015 	/* Driver Statistics */
4016 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "dropped",
4017 			CTLFLAG_RD, &adapter->dropped_pkts,
4018 			"Driver dropped packets");
4019 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "link_irq",
4020 			CTLFLAG_RD, &adapter->link_irq,
4021 			"Link MSI-X IRQ Handled");
4022 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "rx_overruns",
4023 			CTLFLAG_RD, &adapter->rx_overruns,
4024 			"RX overruns");
4025 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "watchdog_timeouts",
4026 			CTLFLAG_RD, &adapter->watchdog_events,
4027 			"Watchdog timeouts");
4028 	SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "device_control",
4029 			CTLTYPE_UINT | CTLFLAG_RD, adapter, E1000_CTRL,
4030 			em_sysctl_reg_handler, "IU",
4031 			"Device Control Register");
4032 	SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "rx_control",
4033 			CTLTYPE_UINT | CTLFLAG_RD, adapter, E1000_RCTL,
4034 			em_sysctl_reg_handler, "IU",
4035 			"Receiver Control Register");
4036 	SYSCTL_ADD_UINT(ctx, child, OID_AUTO, "fc_high_water",
4037 			CTLFLAG_RD, &adapter->hw.fc.high_water, 0,
4038 			"Flow Control High Watermark");
4039 	SYSCTL_ADD_UINT(ctx, child, OID_AUTO, "fc_low_water",
4040 			CTLFLAG_RD, &adapter->hw.fc.low_water, 0,
4041 			"Flow Control Low Watermark");
4042 
4043 	for (int i = 0; i < adapter->tx_num_queues; i++, tx_que++) {
4044 		struct tx_ring *txr = &tx_que->txr;
4045 		snprintf(namebuf, QUEUE_NAME_LEN, "queue_tx_%d", i);
4046 		queue_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, namebuf,
4047 					    CTLFLAG_RD, NULL, "TX Queue Name");
4048 		queue_list = SYSCTL_CHILDREN(queue_node);
4049 
4050 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "txd_head",
4051 				CTLTYPE_UINT | CTLFLAG_RD, adapter,
4052 				E1000_TDH(txr->me),
4053 				em_sysctl_reg_handler, "IU",
4054 				"Transmit Descriptor Head");
4055 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "txd_tail",
4056 				CTLTYPE_UINT | CTLFLAG_RD, adapter,
4057 				E1000_TDT(txr->me),
4058 				em_sysctl_reg_handler, "IU",
4059 				"Transmit Descriptor Tail");
4060 		SYSCTL_ADD_ULONG(ctx, queue_list, OID_AUTO, "tx_irq",
4061 				CTLFLAG_RD, &txr->tx_irq,
4062 				"Queue MSI-X Transmit Interrupts");
4063 	}
4064 
4065 	for (int j = 0; j < adapter->rx_num_queues; j++, rx_que++) {
4066 		struct rx_ring *rxr = &rx_que->rxr;
4067 		snprintf(namebuf, QUEUE_NAME_LEN, "queue_rx_%d", j);
4068 		queue_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, namebuf,
4069 					    CTLFLAG_RD, NULL, "RX Queue Name");
4070 		queue_list = SYSCTL_CHILDREN(queue_node);
4071 
4072 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "rxd_head",
4073 				CTLTYPE_UINT | CTLFLAG_RD, adapter,
4074 				E1000_RDH(rxr->me),
4075 				em_sysctl_reg_handler, "IU",
4076 				"Receive Descriptor Head");
4077 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "rxd_tail",
4078 				CTLTYPE_UINT | CTLFLAG_RD, adapter,
4079 				E1000_RDT(rxr->me),
4080 				em_sysctl_reg_handler, "IU",
4081 				"Receive Descriptor Tail");
4082 		SYSCTL_ADD_ULONG(ctx, queue_list, OID_AUTO, "rx_irq",
4083 				CTLFLAG_RD, &rxr->rx_irq,
4084 				"Queue MSI-X Receive Interrupts");
4085 	}
4086 
4087 	/* MAC stats get their own sub node */
4088 
4089 	stat_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, "mac_stats",
4090 				    CTLFLAG_RD, NULL, "Statistics");
4091 	stat_list = SYSCTL_CHILDREN(stat_node);
4092 
4093 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "excess_coll",
4094 			CTLFLAG_RD, &stats->ecol,
4095 			"Excessive collisions");
4096 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "single_coll",
4097 			CTLFLAG_RD, &stats->scc,
4098 			"Single collisions");
4099 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "multiple_coll",
4100 			CTLFLAG_RD, &stats->mcc,
4101 			"Multiple collisions");
4102 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "late_coll",
4103 			CTLFLAG_RD, &stats->latecol,
4104 			"Late collisions");
4105 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "collision_count",
4106 			CTLFLAG_RD, &stats->colc,
4107 			"Collision Count");
4108 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "symbol_errors",
4109 			CTLFLAG_RD, &adapter->stats.symerrs,
4110 			"Symbol Errors");
4111 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "sequence_errors",
4112 			CTLFLAG_RD, &adapter->stats.sec,
4113 			"Sequence Errors");
4114 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "defer_count",
4115 			CTLFLAG_RD, &adapter->stats.dc,
4116 			"Defer Count");
4117 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "missed_packets",
4118 			CTLFLAG_RD, &adapter->stats.mpc,
4119 			"Missed Packets");
4120 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_no_buff",
4121 			CTLFLAG_RD, &adapter->stats.rnbc,
4122 			"Receive No Buffers");
4123 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_undersize",
4124 			CTLFLAG_RD, &adapter->stats.ruc,
4125 			"Receive Undersize");
4126 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_fragmented",
4127 			CTLFLAG_RD, &adapter->stats.rfc,
4128 			"Fragmented Packets Received ");
4129 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_oversize",
4130 			CTLFLAG_RD, &adapter->stats.roc,
4131 			"Oversized Packets Received");
4132 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_jabber",
4133 			CTLFLAG_RD, &adapter->stats.rjc,
4134 			"Recevied Jabber");
4135 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_errs",
4136 			CTLFLAG_RD, &adapter->stats.rxerrc,
4137 			"Receive Errors");
4138 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "crc_errs",
4139 			CTLFLAG_RD, &adapter->stats.crcerrs,
4140 			"CRC errors");
4141 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "alignment_errs",
4142 			CTLFLAG_RD, &adapter->stats.algnerrc,
4143 			"Alignment Errors");
4144 	/* On 82575 these are collision counts */
4145 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "coll_ext_errs",
4146 			CTLFLAG_RD, &adapter->stats.cexterr,
4147 			"Collision/Carrier extension errors");
4148 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xon_recvd",
4149 			CTLFLAG_RD, &adapter->stats.xonrxc,
4150 			"XON Received");
4151 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xon_txd",
4152 			CTLFLAG_RD, &adapter->stats.xontxc,
4153 			"XON Transmitted");
4154 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xoff_recvd",
4155 			CTLFLAG_RD, &adapter->stats.xoffrxc,
4156 			"XOFF Received");
4157 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xoff_txd",
4158 			CTLFLAG_RD, &adapter->stats.xofftxc,
4159 			"XOFF Transmitted");
4160 
4161 	/* Packet Reception Stats */
4162 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_pkts_recvd",
4163 			CTLFLAG_RD, &adapter->stats.tpr,
4164 			"Total Packets Received ");
4165 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_pkts_recvd",
4166 			CTLFLAG_RD, &adapter->stats.gprc,
4167 			"Good Packets Received");
4168 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "bcast_pkts_recvd",
4169 			CTLFLAG_RD, &adapter->stats.bprc,
4170 			"Broadcast Packets Received");
4171 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "mcast_pkts_recvd",
4172 			CTLFLAG_RD, &adapter->stats.mprc,
4173 			"Multicast Packets Received");
4174 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_64",
4175 			CTLFLAG_RD, &adapter->stats.prc64,
4176 			"64 byte frames received ");
4177 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_65_127",
4178 			CTLFLAG_RD, &adapter->stats.prc127,
4179 			"65-127 byte frames received");
4180 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_128_255",
4181 			CTLFLAG_RD, &adapter->stats.prc255,
4182 			"128-255 byte frames received");
4183 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_256_511",
4184 			CTLFLAG_RD, &adapter->stats.prc511,
4185 			"256-511 byte frames received");
4186 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_512_1023",
4187 			CTLFLAG_RD, &adapter->stats.prc1023,
4188 			"512-1023 byte frames received");
4189 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_1024_1522",
4190 			CTLFLAG_RD, &adapter->stats.prc1522,
4191 			"1023-1522 byte frames received");
4192 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_octets_recvd",
4193 			CTLFLAG_RD, &adapter->stats.gorc,
4194 			"Good Octets Received");
4195 
4196 	/* Packet Transmission Stats */
4197 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_octets_txd",
4198 			CTLFLAG_RD, &adapter->stats.gotc,
4199 			"Good Octets Transmitted");
4200 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_pkts_txd",
4201 			CTLFLAG_RD, &adapter->stats.tpt,
4202 			"Total Packets Transmitted");
4203 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_pkts_txd",
4204 			CTLFLAG_RD, &adapter->stats.gptc,
4205 			"Good Packets Transmitted");
4206 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "bcast_pkts_txd",
4207 			CTLFLAG_RD, &adapter->stats.bptc,
4208 			"Broadcast Packets Transmitted");
4209 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "mcast_pkts_txd",
4210 			CTLFLAG_RD, &adapter->stats.mptc,
4211 			"Multicast Packets Transmitted");
4212 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_64",
4213 			CTLFLAG_RD, &adapter->stats.ptc64,
4214 			"64 byte frames transmitted ");
4215 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_65_127",
4216 			CTLFLAG_RD, &adapter->stats.ptc127,
4217 			"65-127 byte frames transmitted");
4218 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_128_255",
4219 			CTLFLAG_RD, &adapter->stats.ptc255,
4220 			"128-255 byte frames transmitted");
4221 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_256_511",
4222 			CTLFLAG_RD, &adapter->stats.ptc511,
4223 			"256-511 byte frames transmitted");
4224 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_512_1023",
4225 			CTLFLAG_RD, &adapter->stats.ptc1023,
4226 			"512-1023 byte frames transmitted");
4227 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_1024_1522",
4228 			CTLFLAG_RD, &adapter->stats.ptc1522,
4229 			"1024-1522 byte frames transmitted");
4230 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tso_txd",
4231 			CTLFLAG_RD, &adapter->stats.tsctc,
4232 			"TSO Contexts Transmitted");
4233 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tso_ctx_fail",
4234 			CTLFLAG_RD, &adapter->stats.tsctfc,
4235 			"TSO Contexts Failed");
4236 
4237 
4238 	/* Interrupt Stats */
4239 
4240 	int_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, "interrupts",
4241 				    CTLFLAG_RD, NULL, "Interrupt Statistics");
4242 	int_list = SYSCTL_CHILDREN(int_node);
4243 
4244 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "asserts",
4245 			CTLFLAG_RD, &adapter->stats.iac,
4246 			"Interrupt Assertion Count");
4247 
4248 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_pkt_timer",
4249 			CTLFLAG_RD, &adapter->stats.icrxptc,
4250 			"Interrupt Cause Rx Pkt Timer Expire Count");
4251 
4252 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_abs_timer",
4253 			CTLFLAG_RD, &adapter->stats.icrxatc,
4254 			"Interrupt Cause Rx Abs Timer Expire Count");
4255 
4256 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_pkt_timer",
4257 			CTLFLAG_RD, &adapter->stats.ictxptc,
4258 			"Interrupt Cause Tx Pkt Timer Expire Count");
4259 
4260 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_abs_timer",
4261 			CTLFLAG_RD, &adapter->stats.ictxatc,
4262 			"Interrupt Cause Tx Abs Timer Expire Count");
4263 
4264 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_queue_empty",
4265 			CTLFLAG_RD, &adapter->stats.ictxqec,
4266 			"Interrupt Cause Tx Queue Empty Count");
4267 
4268 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_queue_min_thresh",
4269 			CTLFLAG_RD, &adapter->stats.ictxqmtc,
4270 			"Interrupt Cause Tx Queue Min Thresh Count");
4271 
4272 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_desc_min_thresh",
4273 			CTLFLAG_RD, &adapter->stats.icrxdmtc,
4274 			"Interrupt Cause Rx Desc Min Thresh Count");
4275 
4276 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_overrun",
4277 			CTLFLAG_RD, &adapter->stats.icrxoc,
4278 			"Interrupt Cause Receiver Overrun Count");
4279 }
4280 
4281 /**********************************************************************
4282  *
4283  *  This routine provides a way to dump out the adapter eeprom,
4284  *  often a useful debug/service tool. This only dumps the first
4285  *  32 words, stuff that matters is in that extent.
4286  *
4287  **********************************************************************/
4288 static int
4289 em_sysctl_nvm_info(SYSCTL_HANDLER_ARGS)
4290 {
4291 	struct adapter *adapter = (struct adapter *)arg1;
4292 	int error;
4293 	int result;
4294 
4295 	result = -1;
4296 	error = sysctl_handle_int(oidp, &result, 0, req);
4297 
4298 	if (error || !req->newptr)
4299 		return (error);
4300 
4301 	/*
4302 	 * This value will cause a hex dump of the
4303 	 * first 32 16-bit words of the EEPROM to
4304 	 * the screen.
4305 	 */
4306 	if (result == 1)
4307 		em_print_nvm_info(adapter);
4308 
4309 	return (error);
4310 }
4311 
4312 static void
4313 em_print_nvm_info(struct adapter *adapter)
4314 {
4315 	u16 eeprom_data;
4316 	int i, j, row = 0;
4317 
4318 	/* Its a bit crude, but it gets the job done */
4319 	printf("\nInterface EEPROM Dump:\n");
4320 	printf("Offset\n0x0000  ");
4321 	for (i = 0, j = 0; i < 32; i++, j++) {
4322 		if (j == 8) { /* Make the offset block */
4323 			j = 0; ++row;
4324 			printf("\n0x00%x0  ",row);
4325 		}
4326 		e1000_read_nvm(&adapter->hw, i, 1, &eeprom_data);
4327 		printf("%04x ", eeprom_data);
4328 	}
4329 	printf("\n");
4330 }
4331 
4332 static int
4333 em_sysctl_int_delay(SYSCTL_HANDLER_ARGS)
4334 {
4335 	struct em_int_delay_info *info;
4336 	struct adapter *adapter;
4337 	u32 regval;
4338 	int error, usecs, ticks;
4339 
4340 	info = (struct em_int_delay_info *) arg1;
4341 	usecs = info->value;
4342 	error = sysctl_handle_int(oidp, &usecs, 0, req);
4343 	if (error != 0 || req->newptr == NULL)
4344 		return (error);
4345 	if (usecs < 0 || usecs > EM_TICKS_TO_USECS(65535))
4346 		return (EINVAL);
4347 	info->value = usecs;
4348 	ticks = EM_USECS_TO_TICKS(usecs);
4349 	if (info->offset == E1000_ITR)	/* units are 256ns here */
4350 		ticks *= 4;
4351 
4352 	adapter = info->adapter;
4353 
4354 	regval = E1000_READ_OFFSET(&adapter->hw, info->offset);
4355 	regval = (regval & ~0xffff) | (ticks & 0xffff);
4356 	/* Handle a few special cases. */
4357 	switch (info->offset) {
4358 	case E1000_RDTR:
4359 		break;
4360 	case E1000_TIDV:
4361 		if (ticks == 0) {
4362 			adapter->txd_cmd &= ~E1000_TXD_CMD_IDE;
4363 			/* Don't write 0 into the TIDV register. */
4364 			regval++;
4365 		} else
4366 			adapter->txd_cmd |= E1000_TXD_CMD_IDE;
4367 		break;
4368 	}
4369 	E1000_WRITE_OFFSET(&adapter->hw, info->offset, regval);
4370 	return (0);
4371 }
4372 
4373 static void
4374 em_add_int_delay_sysctl(struct adapter *adapter, const char *name,
4375 	const char *description, struct em_int_delay_info *info,
4376 	int offset, int value)
4377 {
4378 	info->adapter = adapter;
4379 	info->offset = offset;
4380 	info->value = value;
4381 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(adapter->dev),
4382 	    SYSCTL_CHILDREN(device_get_sysctl_tree(adapter->dev)),
4383 	    OID_AUTO, name, CTLTYPE_INT|CTLFLAG_RW,
4384 	    info, 0, em_sysctl_int_delay, "I", description);
4385 }
4386 
4387 /*
4388  * Set flow control using sysctl:
4389  * Flow control values:
4390  *      0 - off
4391  *      1 - rx pause
4392  *      2 - tx pause
4393  *      3 - full
4394  */
4395 static int
4396 em_set_flowcntl(SYSCTL_HANDLER_ARGS)
4397 {
4398 	int error;
4399 	static int input = 3; /* default is full */
4400 	struct adapter	*adapter = (struct adapter *) arg1;
4401 
4402 	error = sysctl_handle_int(oidp, &input, 0, req);
4403 
4404 	if ((error) || (req->newptr == NULL))
4405 		return (error);
4406 
4407 	if (input == adapter->fc) /* no change? */
4408 		return (error);
4409 
4410 	switch (input) {
4411 	case e1000_fc_rx_pause:
4412 	case e1000_fc_tx_pause:
4413 	case e1000_fc_full:
4414 	case e1000_fc_none:
4415 		adapter->hw.fc.requested_mode = input;
4416 		adapter->fc = input;
4417 		break;
4418 	default:
4419 		/* Do nothing */
4420 		return (error);
4421 	}
4422 
4423 	adapter->hw.fc.current_mode = adapter->hw.fc.requested_mode;
4424 	e1000_force_mac_fc(&adapter->hw);
4425 	return (error);
4426 }
4427 
4428 /*
4429  * Manage Energy Efficient Ethernet:
4430  * Control values:
4431  *     0/1 - enabled/disabled
4432  */
4433 static int
4434 em_sysctl_eee(SYSCTL_HANDLER_ARGS)
4435 {
4436 	struct adapter *adapter = (struct adapter *) arg1;
4437 	int error, value;
4438 
4439 	value = adapter->hw.dev_spec.ich8lan.eee_disable;
4440 	error = sysctl_handle_int(oidp, &value, 0, req);
4441 	if (error || req->newptr == NULL)
4442 		return (error);
4443 	adapter->hw.dev_spec.ich8lan.eee_disable = (value != 0);
4444 	em_if_init(adapter->ctx);
4445 
4446 	return (0);
4447 }
4448 
4449 static int
4450 em_sysctl_debug_info(SYSCTL_HANDLER_ARGS)
4451 {
4452 	struct adapter *adapter;
4453 	int error;
4454 	int result;
4455 
4456 	result = -1;
4457 	error = sysctl_handle_int(oidp, &result, 0, req);
4458 
4459 	if (error || !req->newptr)
4460 		return (error);
4461 
4462 	if (result == 1) {
4463 		adapter = (struct adapter *) arg1;
4464 		em_print_debug_info(adapter);
4465 	}
4466 
4467 	return (error);
4468 }
4469 
4470 static int
4471 em_get_rs(SYSCTL_HANDLER_ARGS)
4472 {
4473 	struct adapter *adapter = (struct adapter *) arg1;
4474 	int error;
4475 	int result;
4476 
4477 	result = 0;
4478 	error = sysctl_handle_int(oidp, &result, 0, req);
4479 
4480 	if (error || !req->newptr || result != 1)
4481 		return (error);
4482 	em_dump_rs(adapter);
4483 
4484 	return (error);
4485 }
4486 
4487 static void
4488 em_if_debug(if_ctx_t ctx)
4489 {
4490 	em_dump_rs(iflib_get_softc(ctx));
4491 }
4492 
4493 /*
4494  * This routine is meant to be fluid, add whatever is
4495  * needed for debugging a problem.  -jfv
4496  */
4497 static void
4498 em_print_debug_info(struct adapter *adapter)
4499 {
4500 	device_t dev = iflib_get_dev(adapter->ctx);
4501 	struct ifnet *ifp = iflib_get_ifp(adapter->ctx);
4502 	struct tx_ring *txr = &adapter->tx_queues->txr;
4503 	struct rx_ring *rxr = &adapter->rx_queues->rxr;
4504 
4505 	if (if_getdrvflags(ifp) & IFF_DRV_RUNNING)
4506 		printf("Interface is RUNNING ");
4507 	else
4508 		printf("Interface is NOT RUNNING\n");
4509 
4510 	if (if_getdrvflags(ifp) & IFF_DRV_OACTIVE)
4511 		printf("and INACTIVE\n");
4512 	else
4513 		printf("and ACTIVE\n");
4514 
4515 	for (int i = 0; i < adapter->tx_num_queues; i++, txr++) {
4516 		device_printf(dev, "TX Queue %d ------\n", i);
4517 		device_printf(dev, "hw tdh = %d, hw tdt = %d\n",
4518 			E1000_READ_REG(&adapter->hw, E1000_TDH(i)),
4519 			E1000_READ_REG(&adapter->hw, E1000_TDT(i)));
4520 
4521 	}
4522 	for (int j=0; j < adapter->rx_num_queues; j++, rxr++) {
4523 		device_printf(dev, "RX Queue %d ------\n", j);
4524 		device_printf(dev, "hw rdh = %d, hw rdt = %d\n",
4525 			E1000_READ_REG(&adapter->hw, E1000_RDH(j)),
4526 			E1000_READ_REG(&adapter->hw, E1000_RDT(j)));
4527 	}
4528 }
4529 
4530 /*
4531  * 82574 only:
4532  * Write a new value to the EEPROM increasing the number of MSI-X
4533  * vectors from 3 to 5, for proper multiqueue support.
4534  */
4535 static void
4536 em_enable_vectors_82574(if_ctx_t ctx)
4537 {
4538 	struct adapter *adapter = iflib_get_softc(ctx);
4539 	struct e1000_hw *hw = &adapter->hw;
4540 	device_t dev = iflib_get_dev(ctx);
4541 	u16 edata;
4542 
4543 	e1000_read_nvm(hw, EM_NVM_PCIE_CTRL, 1, &edata);
4544 	if (bootverbose)
4545 		device_printf(dev, "EM_NVM_PCIE_CTRL = %#06x\n", edata);
4546 	if (((edata & EM_NVM_MSIX_N_MASK) >> EM_NVM_MSIX_N_SHIFT) != 4) {
4547 		device_printf(dev, "Writing to eeprom: increasing "
4548 		    "reported MSI-X vectors from 3 to 5...\n");
4549 		edata &= ~(EM_NVM_MSIX_N_MASK);
4550 		edata |= 4 << EM_NVM_MSIX_N_SHIFT;
4551 		e1000_write_nvm(hw, EM_NVM_PCIE_CTRL, 1, &edata);
4552 		e1000_update_nvm_checksum(hw);
4553 		device_printf(dev, "Writing to eeprom: done\n");
4554 	}
4555 }
4556