xref: /freebsd/sys/dev/e1000/if_em.c (revision d0ba1baed3f6e4936a0c1b89c25f6c59168ef6de)
1 /*-
2  * SPDX-License-Identifier: BSD-2-Clause
3  *
4  * Copyright (c) 2016 Nicole Graziano <nicole@nextbsd.org>
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  *
16  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
17  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
20  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
26  * SUCH DAMAGE.
27  */
28 
29 /* $FreeBSD$ */
30 #include "if_em.h"
31 #include <sys/sbuf.h>
32 #include <machine/_inttypes.h>
33 
34 #define em_mac_min e1000_82547
35 #define igb_mac_min e1000_82575
36 
37 /*********************************************************************
38  *  Driver version:
39  *********************************************************************/
40 char em_driver_version[] = "7.6.1-k";
41 
42 /*********************************************************************
43  *  PCI Device ID Table
44  *
45  *  Used by probe to select devices to load on
46  *  Last field stores an index into e1000_strings
47  *  Last entry must be all 0s
48  *
49  *  { Vendor ID, Device ID, SubVendor ID, SubDevice ID, String Index }
50  *********************************************************************/
51 
52 static pci_vendor_info_t em_vendor_info_array[] =
53 {
54 	/* Intel(R) PRO/1000 Network Connection - Legacy em*/
55 	PVID(0x8086, E1000_DEV_ID_82540EM, "Intel(R) PRO/1000 Network Connection"),
56 	PVID(0x8086, E1000_DEV_ID_82540EM_LOM, "Intel(R) PRO/1000 Network Connection"),
57 	PVID(0x8086, E1000_DEV_ID_82540EP, "Intel(R) PRO/1000 Network Connection"),
58 	PVID(0x8086, E1000_DEV_ID_82540EP_LOM, "Intel(R) PRO/1000 Network Connection"),
59 	PVID(0x8086, E1000_DEV_ID_82540EP_LP, "Intel(R) PRO/1000 Network Connection"),
60 
61 	PVID(0x8086, E1000_DEV_ID_82541EI, "Intel(R) PRO/1000 Network Connection"),
62 	PVID(0x8086, E1000_DEV_ID_82541ER, "Intel(R) PRO/1000 Network Connection"),
63 	PVID(0x8086, E1000_DEV_ID_82541ER_LOM, "Intel(R) PRO/1000 Network Connection"),
64 	PVID(0x8086, E1000_DEV_ID_82541EI_MOBILE, "Intel(R) PRO/1000 Network Connection"),
65 	PVID(0x8086, E1000_DEV_ID_82541GI, "Intel(R) PRO/1000 Network Connection"),
66 	PVID(0x8086, E1000_DEV_ID_82541GI_LF, "Intel(R) PRO/1000 Network Connection"),
67 	PVID(0x8086, E1000_DEV_ID_82541GI_MOBILE, "Intel(R) PRO/1000 Network Connection"),
68 
69 	PVID(0x8086, E1000_DEV_ID_82542, "Intel(R) PRO/1000 Network Connection"),
70 
71 	PVID(0x8086, E1000_DEV_ID_82543GC_FIBER, "Intel(R) PRO/1000 Network Connection"),
72 	PVID(0x8086, E1000_DEV_ID_82543GC_COPPER, "Intel(R) PRO/1000 Network Connection"),
73 
74 	PVID(0x8086, E1000_DEV_ID_82544EI_COPPER, "Intel(R) PRO/1000 Network Connection"),
75 	PVID(0x8086, E1000_DEV_ID_82544EI_FIBER, "Intel(R) PRO/1000 Network Connection"),
76 	PVID(0x8086, E1000_DEV_ID_82544GC_COPPER, "Intel(R) PRO/1000 Network Connection"),
77 	PVID(0x8086, E1000_DEV_ID_82544GC_LOM, "Intel(R) PRO/1000 Network Connection"),
78 
79 	PVID(0x8086, E1000_DEV_ID_82545EM_COPPER, "Intel(R) PRO/1000 Network Connection"),
80 	PVID(0x8086, E1000_DEV_ID_82545EM_FIBER, "Intel(R) PRO/1000 Network Connection"),
81 	PVID(0x8086, E1000_DEV_ID_82545GM_COPPER, "Intel(R) PRO/1000 Network Connection"),
82 	PVID(0x8086, E1000_DEV_ID_82545GM_FIBER, "Intel(R) PRO/1000 Network Connection"),
83 	PVID(0x8086, E1000_DEV_ID_82545GM_SERDES, "Intel(R) PRO/1000 Network Connection"),
84 
85 	PVID(0x8086, E1000_DEV_ID_82546EB_COPPER, "Intel(R) PRO/1000 Network Connection"),
86 	PVID(0x8086, E1000_DEV_ID_82546EB_FIBER, "Intel(R) PRO/1000 Network Connection"),
87 	PVID(0x8086, E1000_DEV_ID_82546EB_QUAD_COPPER, "Intel(R) PRO/1000 Network Connection"),
88 	PVID(0x8086, E1000_DEV_ID_82546GB_COPPER, "Intel(R) PRO/1000 Network Connection"),
89 	PVID(0x8086, E1000_DEV_ID_82546GB_FIBER, "Intel(R) PRO/1000 Network Connection"),
90 	PVID(0x8086, E1000_DEV_ID_82546GB_SERDES, "Intel(R) PRO/1000 Network Connection"),
91 	PVID(0x8086, E1000_DEV_ID_82546GB_PCIE, "Intel(R) PRO/1000 Network Connection"),
92 	PVID(0x8086, E1000_DEV_ID_82546GB_QUAD_COPPER, "Intel(R) PRO/1000 Network Connection"),
93 	PVID(0x8086, E1000_DEV_ID_82546GB_QUAD_COPPER_KSP3, "Intel(R) PRO/1000 Network Connection"),
94 
95 	PVID(0x8086, E1000_DEV_ID_82547EI, "Intel(R) PRO/1000 Network Connection"),
96 	PVID(0x8086, E1000_DEV_ID_82547EI_MOBILE, "Intel(R) PRO/1000 Network Connection"),
97 	PVID(0x8086, E1000_DEV_ID_82547GI, "Intel(R) PRO/1000 Network Connection"),
98 
99 	/* Intel(R) PRO/1000 Network Connection - em */
100 	PVID(0x8086, E1000_DEV_ID_82571EB_COPPER, "Intel(R) PRO/1000 Network Connection"),
101 	PVID(0x8086, E1000_DEV_ID_82571EB_FIBER, "Intel(R) PRO/1000 Network Connection"),
102 	PVID(0x8086, E1000_DEV_ID_82571EB_SERDES, "Intel(R) PRO/1000 Network Connection"),
103 	PVID(0x8086, E1000_DEV_ID_82571EB_SERDES_DUAL, "Intel(R) PRO/1000 Network Connection"),
104 	PVID(0x8086, E1000_DEV_ID_82571EB_SERDES_QUAD, "Intel(R) PRO/1000 Network Connection"),
105 	PVID(0x8086, E1000_DEV_ID_82571EB_QUAD_COPPER, "Intel(R) PRO/1000 Network Connection"),
106 	PVID(0x8086, E1000_DEV_ID_82571EB_QUAD_COPPER_LP, "Intel(R) PRO/1000 Network Connection"),
107 	PVID(0x8086, E1000_DEV_ID_82571EB_QUAD_FIBER, "Intel(R) PRO/1000 Network Connection"),
108 	PVID(0x8086, E1000_DEV_ID_82571PT_QUAD_COPPER, "Intel(R) PRO/1000 Network Connection"),
109 	PVID(0x8086, E1000_DEV_ID_82572EI, "Intel(R) PRO/1000 Network Connection"),
110 	PVID(0x8086, E1000_DEV_ID_82572EI_COPPER, "Intel(R) PRO/1000 Network Connection"),
111 	PVID(0x8086, E1000_DEV_ID_82572EI_FIBER, "Intel(R) PRO/1000 Network Connection"),
112 	PVID(0x8086, E1000_DEV_ID_82572EI_SERDES, "Intel(R) PRO/1000 Network Connection"),
113 	PVID(0x8086, E1000_DEV_ID_82573E, "Intel(R) PRO/1000 Network Connection"),
114 	PVID(0x8086, E1000_DEV_ID_82573E_IAMT, "Intel(R) PRO/1000 Network Connection"),
115 	PVID(0x8086, E1000_DEV_ID_82573L, "Intel(R) PRO/1000 Network Connection"),
116 	PVID(0x8086, E1000_DEV_ID_82583V, "Intel(R) PRO/1000 Network Connection"),
117 	PVID(0x8086, E1000_DEV_ID_80003ES2LAN_COPPER_SPT, "Intel(R) PRO/1000 Network Connection"),
118 	PVID(0x8086, E1000_DEV_ID_80003ES2LAN_SERDES_SPT, "Intel(R) PRO/1000 Network Connection"),
119 	PVID(0x8086, E1000_DEV_ID_80003ES2LAN_COPPER_DPT, "Intel(R) PRO/1000 Network Connection"),
120 	PVID(0x8086, E1000_DEV_ID_80003ES2LAN_SERDES_DPT, "Intel(R) PRO/1000 Network Connection"),
121 	PVID(0x8086, E1000_DEV_ID_ICH8_IGP_M_AMT, "Intel(R) PRO/1000 Network Connection"),
122 	PVID(0x8086, E1000_DEV_ID_ICH8_IGP_AMT, "Intel(R) PRO/1000 Network Connection"),
123 	PVID(0x8086, E1000_DEV_ID_ICH8_IGP_C, "Intel(R) PRO/1000 Network Connection"),
124 	PVID(0x8086, E1000_DEV_ID_ICH8_IFE, "Intel(R) PRO/1000 Network Connection"),
125 	PVID(0x8086, E1000_DEV_ID_ICH8_IFE_GT, "Intel(R) PRO/1000 Network Connection"),
126 	PVID(0x8086, E1000_DEV_ID_ICH8_IFE_G, "Intel(R) PRO/1000 Network Connection"),
127 	PVID(0x8086, E1000_DEV_ID_ICH8_IGP_M, "Intel(R) PRO/1000 Network Connection"),
128 	PVID(0x8086, E1000_DEV_ID_ICH8_82567V_3, "Intel(R) PRO/1000 Network Connection"),
129 	PVID(0x8086, E1000_DEV_ID_ICH9_IGP_M_AMT, "Intel(R) PRO/1000 Network Connection"),
130 	PVID(0x8086, E1000_DEV_ID_ICH9_IGP_AMT, "Intel(R) PRO/1000 Network Connection"),
131 	PVID(0x8086, E1000_DEV_ID_ICH9_IGP_C, "Intel(R) PRO/1000 Network Connection"),
132 	PVID(0x8086, E1000_DEV_ID_ICH9_IGP_M, "Intel(R) PRO/1000 Network Connection"),
133 	PVID(0x8086, E1000_DEV_ID_ICH9_IGP_M_V, "Intel(R) PRO/1000 Network Connection"),
134 	PVID(0x8086, E1000_DEV_ID_ICH9_IFE, "Intel(R) PRO/1000 Network Connection"),
135 	PVID(0x8086, E1000_DEV_ID_ICH9_IFE_GT, "Intel(R) PRO/1000 Network Connection"),
136 	PVID(0x8086, E1000_DEV_ID_ICH9_IFE_G, "Intel(R) PRO/1000 Network Connection"),
137 	PVID(0x8086, E1000_DEV_ID_ICH9_BM, "Intel(R) PRO/1000 Network Connection"),
138 	PVID(0x8086, E1000_DEV_ID_82574L, "Intel(R) PRO/1000 Network Connection"),
139 	PVID(0x8086, E1000_DEV_ID_82574LA, "Intel(R) PRO/1000 Network Connection"),
140 	PVID(0x8086, E1000_DEV_ID_ICH10_R_BM_LM, "Intel(R) PRO/1000 Network Connection"),
141 	PVID(0x8086, E1000_DEV_ID_ICH10_R_BM_LF, "Intel(R) PRO/1000 Network Connection"),
142 	PVID(0x8086, E1000_DEV_ID_ICH10_R_BM_V, "Intel(R) PRO/1000 Network Connection"),
143 	PVID(0x8086, E1000_DEV_ID_ICH10_D_BM_LM, "Intel(R) PRO/1000 Network Connection"),
144 	PVID(0x8086, E1000_DEV_ID_ICH10_D_BM_LF, "Intel(R) PRO/1000 Network Connection"),
145 	PVID(0x8086, E1000_DEV_ID_ICH10_D_BM_V, "Intel(R) PRO/1000 Network Connection"),
146 	PVID(0x8086, E1000_DEV_ID_PCH_M_HV_LM, "Intel(R) PRO/1000 Network Connection"),
147 	PVID(0x8086, E1000_DEV_ID_PCH_M_HV_LC, "Intel(R) PRO/1000 Network Connection"),
148 	PVID(0x8086, E1000_DEV_ID_PCH_D_HV_DM, "Intel(R) PRO/1000 Network Connection"),
149 	PVID(0x8086, E1000_DEV_ID_PCH_D_HV_DC, "Intel(R) PRO/1000 Network Connection"),
150 	PVID(0x8086, E1000_DEV_ID_PCH2_LV_LM, "Intel(R) PRO/1000 Network Connection"),
151 	PVID(0x8086, E1000_DEV_ID_PCH2_LV_V, "Intel(R) PRO/1000 Network Connection"),
152 	PVID(0x8086, E1000_DEV_ID_PCH_LPT_I217_LM, "Intel(R) PRO/1000 Network Connection"),
153 	PVID(0x8086, E1000_DEV_ID_PCH_LPT_I217_V, "Intel(R) PRO/1000 Network Connection"),
154 	PVID(0x8086, E1000_DEV_ID_PCH_LPTLP_I218_LM, "Intel(R) PRO/1000 Network Connection"),
155 	PVID(0x8086, E1000_DEV_ID_PCH_LPTLP_I218_V, "Intel(R) PRO/1000 Network Connection"),
156 	PVID(0x8086, E1000_DEV_ID_PCH_I218_LM2, "Intel(R) PRO/1000 Network Connection"),
157 	PVID(0x8086, E1000_DEV_ID_PCH_I218_V2, "Intel(R) PRO/1000 Network Connection"),
158 	PVID(0x8086, E1000_DEV_ID_PCH_I218_LM3, "Intel(R) PRO/1000 Network Connection"),
159 	PVID(0x8086, E1000_DEV_ID_PCH_I218_V3, "Intel(R) PRO/1000 Network Connection"),
160 	PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_LM, "Intel(R) PRO/1000 Network Connection"),
161 	PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_V, "Intel(R) PRO/1000 Network Connection"),
162 	PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_LM2, "Intel(R) PRO/1000 Network Connection"),
163 	PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_V2, "Intel(R) PRO/1000 Network Connection"),
164 	PVID(0x8086, E1000_DEV_ID_PCH_LBG_I219_LM3, "Intel(R) PRO/1000 Network Connection"),
165 	PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_LM4, "Intel(R) PRO/1000 Network Connection"),
166 	PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_V4, "Intel(R) PRO/1000 Network Connection"),
167 	PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_LM5, "Intel(R) PRO/1000 Network Connection"),
168 	PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_V5, "Intel(R) PRO/1000 Network Connection"),
169 	PVID(0x8086, E1000_DEV_ID_PCH_CNP_I219_LM6, "Intel(R) PRO/1000 Network Connection"),
170 	PVID(0x8086, E1000_DEV_ID_PCH_CNP_I219_V6, "Intel(R) PRO/1000 Network Connection"),
171 	PVID(0x8086, E1000_DEV_ID_PCH_CNP_I219_LM7, "Intel(R) PRO/1000 Network Connection"),
172 	PVID(0x8086, E1000_DEV_ID_PCH_CNP_I219_V7, "Intel(R) PRO/1000 Network Connection"),
173 	PVID(0x8086, E1000_DEV_ID_PCH_ICP_I219_LM8, "Intel(R) PRO/1000 Network Connection"),
174 	PVID(0x8086, E1000_DEV_ID_PCH_ICP_I219_V8, "Intel(R) PRO/1000 Network Connection"),
175 	PVID(0x8086, E1000_DEV_ID_PCH_ICP_I219_LM9, "Intel(R) PRO/1000 Network Connection"),
176 	PVID(0x8086, E1000_DEV_ID_PCH_ICP_I219_V9, "Intel(R) PRO/1000 Network Connection"),
177 	/* required last entry */
178 	PVID_END
179 };
180 
181 static pci_vendor_info_t igb_vendor_info_array[] =
182 {
183 	/* Intel(R) PRO/1000 Network Connection - igb */
184 	PVID(0x8086, E1000_DEV_ID_82575EB_COPPER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
185 	PVID(0x8086, E1000_DEV_ID_82575EB_FIBER_SERDES, "Intel(R) PRO/1000 PCI-Express Network Driver"),
186 	PVID(0x8086, E1000_DEV_ID_82575GB_QUAD_COPPER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
187 	PVID(0x8086, E1000_DEV_ID_82576, "Intel(R) PRO/1000 PCI-Express Network Driver"),
188 	PVID(0x8086, E1000_DEV_ID_82576_NS, "Intel(R) PRO/1000 PCI-Express Network Driver"),
189 	PVID(0x8086, E1000_DEV_ID_82576_NS_SERDES, "Intel(R) PRO/1000 PCI-Express Network Driver"),
190 	PVID(0x8086, E1000_DEV_ID_82576_FIBER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
191 	PVID(0x8086, E1000_DEV_ID_82576_SERDES, "Intel(R) PRO/1000 PCI-Express Network Driver"),
192 	PVID(0x8086, E1000_DEV_ID_82576_SERDES_QUAD, "Intel(R) PRO/1000 PCI-Express Network Driver"),
193 	PVID(0x8086, E1000_DEV_ID_82576_QUAD_COPPER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
194 	PVID(0x8086, E1000_DEV_ID_82576_QUAD_COPPER_ET2, "Intel(R) PRO/1000 PCI-Express Network Driver"),
195 	PVID(0x8086, E1000_DEV_ID_82576_VF, "Intel(R) PRO/1000 PCI-Express Network Driver"),
196 	PVID(0x8086, E1000_DEV_ID_82580_COPPER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
197 	PVID(0x8086, E1000_DEV_ID_82580_FIBER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
198 	PVID(0x8086, E1000_DEV_ID_82580_SERDES, "Intel(R) PRO/1000 PCI-Express Network Driver"),
199 	PVID(0x8086, E1000_DEV_ID_82580_SGMII, "Intel(R) PRO/1000 PCI-Express Network Driver"),
200 	PVID(0x8086, E1000_DEV_ID_82580_COPPER_DUAL, "Intel(R) PRO/1000 PCI-Express Network Driver"),
201 	PVID(0x8086, E1000_DEV_ID_82580_QUAD_FIBER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
202 	PVID(0x8086, E1000_DEV_ID_DH89XXCC_SERDES, "Intel(R) PRO/1000 PCI-Express Network Driver"),
203 	PVID(0x8086, E1000_DEV_ID_DH89XXCC_SGMII, "Intel(R) PRO/1000 PCI-Express Network Driver"),
204 	PVID(0x8086, E1000_DEV_ID_DH89XXCC_SFP, "Intel(R) PRO/1000 PCI-Express Network Driver"),
205 	PVID(0x8086, E1000_DEV_ID_DH89XXCC_BACKPLANE, "Intel(R) PRO/1000 PCI-Express Network Driver"),
206 	PVID(0x8086, E1000_DEV_ID_I350_COPPER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
207 	PVID(0x8086, E1000_DEV_ID_I350_FIBER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
208 	PVID(0x8086, E1000_DEV_ID_I350_SERDES, "Intel(R) PRO/1000 PCI-Express Network Driver"),
209 	PVID(0x8086, E1000_DEV_ID_I350_SGMII, "Intel(R) PRO/1000 PCI-Express Network Driver"),
210 	PVID(0x8086, E1000_DEV_ID_I350_VF, "Intel(R) PRO/1000 PCI-Express Network Driver"),
211 	PVID(0x8086, E1000_DEV_ID_I210_COPPER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
212 	PVID(0x8086, E1000_DEV_ID_I210_COPPER_IT, "Intel(R) PRO/1000 PCI-Express Network Driver"),
213 	PVID(0x8086, E1000_DEV_ID_I210_COPPER_OEM1, "Intel(R) PRO/1000 PCI-Express Network Driver"),
214 	PVID(0x8086, E1000_DEV_ID_I210_COPPER_FLASHLESS, "Intel(R) PRO/1000 PCI-Express Network Driver"),
215 	PVID(0x8086, E1000_DEV_ID_I210_SERDES_FLASHLESS, "Intel(R) PRO/1000 PCI-Express Network Driver"),
216 	PVID(0x8086, E1000_DEV_ID_I210_FIBER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
217 	PVID(0x8086, E1000_DEV_ID_I210_SERDES, "Intel(R) PRO/1000 PCI-Express Network Driver"),
218 	PVID(0x8086, E1000_DEV_ID_I210_SGMII, "Intel(R) PRO/1000 PCI-Express Network Driver"),
219 	PVID(0x8086, E1000_DEV_ID_I211_COPPER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
220 	PVID(0x8086, E1000_DEV_ID_I354_BACKPLANE_1GBPS, "Intel(R) PRO/1000 PCI-Express Network Driver"),
221 	PVID(0x8086, E1000_DEV_ID_I354_BACKPLANE_2_5GBPS, "Intel(R) PRO/1000 PCI-Express Network Driver"),
222 	PVID(0x8086, E1000_DEV_ID_I354_SGMII, "Intel(R) PRO/1000 PCI-Express Network Driver"),
223 	/* required last entry */
224 	PVID_END
225 };
226 
227 /*********************************************************************
228  *  Function prototypes
229  *********************************************************************/
230 static void	*em_register(device_t dev);
231 static void	*igb_register(device_t dev);
232 static int	em_if_attach_pre(if_ctx_t ctx);
233 static int	em_if_attach_post(if_ctx_t ctx);
234 static int	em_if_detach(if_ctx_t ctx);
235 static int	em_if_shutdown(if_ctx_t ctx);
236 static int	em_if_suspend(if_ctx_t ctx);
237 static int	em_if_resume(if_ctx_t ctx);
238 
239 static int	em_if_tx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int ntxqs, int ntxqsets);
240 static int	em_if_rx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int nrxqs, int nrxqsets);
241 static void	em_if_queues_free(if_ctx_t ctx);
242 
243 static uint64_t	em_if_get_counter(if_ctx_t, ift_counter);
244 static void	em_if_init(if_ctx_t ctx);
245 static void	em_if_stop(if_ctx_t ctx);
246 static void	em_if_media_status(if_ctx_t, struct ifmediareq *);
247 static int	em_if_media_change(if_ctx_t ctx);
248 static int	em_if_mtu_set(if_ctx_t ctx, uint32_t mtu);
249 static void	em_if_timer(if_ctx_t ctx, uint16_t qid);
250 static void	em_if_vlan_register(if_ctx_t ctx, u16 vtag);
251 static void	em_if_vlan_unregister(if_ctx_t ctx, u16 vtag);
252 
253 static void	em_identify_hardware(if_ctx_t ctx);
254 static int	em_allocate_pci_resources(if_ctx_t ctx);
255 static void	em_free_pci_resources(if_ctx_t ctx);
256 static void	em_reset(if_ctx_t ctx);
257 static int	em_setup_interface(if_ctx_t ctx);
258 static int	em_setup_msix(if_ctx_t ctx);
259 
260 static void	em_initialize_transmit_unit(if_ctx_t ctx);
261 static void	em_initialize_receive_unit(if_ctx_t ctx);
262 
263 static void	em_if_enable_intr(if_ctx_t ctx);
264 static void	em_if_disable_intr(if_ctx_t ctx);
265 static int	em_if_rx_queue_intr_enable(if_ctx_t ctx, uint16_t rxqid);
266 static int	em_if_tx_queue_intr_enable(if_ctx_t ctx, uint16_t txqid);
267 static void	em_if_multi_set(if_ctx_t ctx);
268 static void	em_if_update_admin_status(if_ctx_t ctx);
269 static void	em_if_debug(if_ctx_t ctx);
270 static void	em_update_stats_counters(struct adapter *);
271 static void	em_add_hw_stats(struct adapter *adapter);
272 static int	em_if_set_promisc(if_ctx_t ctx, int flags);
273 static void	em_setup_vlan_hw_support(struct adapter *);
274 static int	em_sysctl_nvm_info(SYSCTL_HANDLER_ARGS);
275 static void	em_print_nvm_info(struct adapter *);
276 static int	em_sysctl_debug_info(SYSCTL_HANDLER_ARGS);
277 static int	em_get_rs(SYSCTL_HANDLER_ARGS);
278 static void	em_print_debug_info(struct adapter *);
279 static int 	em_is_valid_ether_addr(u8 *);
280 static int	em_sysctl_int_delay(SYSCTL_HANDLER_ARGS);
281 static void	em_add_int_delay_sysctl(struct adapter *, const char *,
282 		    const char *, struct em_int_delay_info *, int, int);
283 /* Management and WOL Support */
284 static void	em_init_manageability(struct adapter *);
285 static void	em_release_manageability(struct adapter *);
286 static void	em_get_hw_control(struct adapter *);
287 static void	em_release_hw_control(struct adapter *);
288 static void	em_get_wakeup(if_ctx_t ctx);
289 static void	em_enable_wakeup(if_ctx_t ctx);
290 static int	em_enable_phy_wakeup(struct adapter *);
291 static void	em_disable_aspm(struct adapter *);
292 
293 int		em_intr(void *arg);
294 static void	em_disable_promisc(if_ctx_t ctx);
295 
296 /* MSIX handlers */
297 static int	em_if_msix_intr_assign(if_ctx_t, int);
298 static int	em_msix_link(void *);
299 static void	em_handle_link(void *context);
300 
301 static void	em_enable_vectors_82574(if_ctx_t);
302 
303 static int	em_set_flowcntl(SYSCTL_HANDLER_ARGS);
304 static int	em_sysctl_eee(SYSCTL_HANDLER_ARGS);
305 static void	em_if_led_func(if_ctx_t ctx, int onoff);
306 
307 static int	em_get_regs(SYSCTL_HANDLER_ARGS);
308 
309 static void	lem_smartspeed(struct adapter *adapter);
310 static void	igb_configure_queues(struct adapter *adapter);
311 
312 
313 /*********************************************************************
314  *  FreeBSD Device Interface Entry Points
315  *********************************************************************/
316 static device_method_t em_methods[] = {
317 	/* Device interface */
318 	DEVMETHOD(device_register, em_register),
319 	DEVMETHOD(device_probe, iflib_device_probe),
320 	DEVMETHOD(device_attach, iflib_device_attach),
321 	DEVMETHOD(device_detach, iflib_device_detach),
322 	DEVMETHOD(device_shutdown, iflib_device_shutdown),
323 	DEVMETHOD(device_suspend, iflib_device_suspend),
324 	DEVMETHOD(device_resume, iflib_device_resume),
325 	DEVMETHOD_END
326 };
327 
328 static device_method_t igb_methods[] = {
329 	/* Device interface */
330 	DEVMETHOD(device_register, igb_register),
331 	DEVMETHOD(device_probe, iflib_device_probe),
332 	DEVMETHOD(device_attach, iflib_device_attach),
333 	DEVMETHOD(device_detach, iflib_device_detach),
334 	DEVMETHOD(device_shutdown, iflib_device_shutdown),
335 	DEVMETHOD(device_suspend, iflib_device_suspend),
336 	DEVMETHOD(device_resume, iflib_device_resume),
337 	DEVMETHOD_END
338 };
339 
340 
341 static driver_t em_driver = {
342 	"em", em_methods, sizeof(struct adapter),
343 };
344 
345 static devclass_t em_devclass;
346 DRIVER_MODULE(em, pci, em_driver, em_devclass, 0, 0);
347 
348 MODULE_DEPEND(em, pci, 1, 1, 1);
349 MODULE_DEPEND(em, ether, 1, 1, 1);
350 MODULE_DEPEND(em, iflib, 1, 1, 1);
351 
352 IFLIB_PNP_INFO(pci, em, em_vendor_info_array);
353 
354 static driver_t igb_driver = {
355 	"igb", igb_methods, sizeof(struct adapter),
356 };
357 
358 static devclass_t igb_devclass;
359 DRIVER_MODULE(igb, pci, igb_driver, igb_devclass, 0, 0);
360 
361 MODULE_DEPEND(igb, pci, 1, 1, 1);
362 MODULE_DEPEND(igb, ether, 1, 1, 1);
363 MODULE_DEPEND(igb, iflib, 1, 1, 1);
364 
365 IFLIB_PNP_INFO(pci, igb, igb_vendor_info_array);
366 
367 static device_method_t em_if_methods[] = {
368 	DEVMETHOD(ifdi_attach_pre, em_if_attach_pre),
369 	DEVMETHOD(ifdi_attach_post, em_if_attach_post),
370 	DEVMETHOD(ifdi_detach, em_if_detach),
371 	DEVMETHOD(ifdi_shutdown, em_if_shutdown),
372 	DEVMETHOD(ifdi_suspend, em_if_suspend),
373 	DEVMETHOD(ifdi_resume, em_if_resume),
374 	DEVMETHOD(ifdi_init, em_if_init),
375 	DEVMETHOD(ifdi_stop, em_if_stop),
376 	DEVMETHOD(ifdi_msix_intr_assign, em_if_msix_intr_assign),
377 	DEVMETHOD(ifdi_intr_enable, em_if_enable_intr),
378 	DEVMETHOD(ifdi_intr_disable, em_if_disable_intr),
379 	DEVMETHOD(ifdi_tx_queues_alloc, em_if_tx_queues_alloc),
380 	DEVMETHOD(ifdi_rx_queues_alloc, em_if_rx_queues_alloc),
381 	DEVMETHOD(ifdi_queues_free, em_if_queues_free),
382 	DEVMETHOD(ifdi_update_admin_status, em_if_update_admin_status),
383 	DEVMETHOD(ifdi_multi_set, em_if_multi_set),
384 	DEVMETHOD(ifdi_media_status, em_if_media_status),
385 	DEVMETHOD(ifdi_media_change, em_if_media_change),
386 	DEVMETHOD(ifdi_mtu_set, em_if_mtu_set),
387 	DEVMETHOD(ifdi_promisc_set, em_if_set_promisc),
388 	DEVMETHOD(ifdi_timer, em_if_timer),
389 	DEVMETHOD(ifdi_vlan_register, em_if_vlan_register),
390 	DEVMETHOD(ifdi_vlan_unregister, em_if_vlan_unregister),
391 	DEVMETHOD(ifdi_get_counter, em_if_get_counter),
392 	DEVMETHOD(ifdi_led_func, em_if_led_func),
393 	DEVMETHOD(ifdi_rx_queue_intr_enable, em_if_rx_queue_intr_enable),
394 	DEVMETHOD(ifdi_tx_queue_intr_enable, em_if_tx_queue_intr_enable),
395 	DEVMETHOD(ifdi_debug, em_if_debug),
396 	DEVMETHOD_END
397 };
398 
399 /*
400  * note that if (adapter->msix_mem) is replaced by:
401  * if (adapter->intr_type == IFLIB_INTR_MSIX)
402  */
403 static driver_t em_if_driver = {
404 	"em_if", em_if_methods, sizeof(struct adapter)
405 };
406 
407 /*********************************************************************
408  *  Tunable default values.
409  *********************************************************************/
410 
411 #define EM_TICKS_TO_USECS(ticks)	((1024 * (ticks) + 500) / 1000)
412 #define EM_USECS_TO_TICKS(usecs)	((1000 * (usecs) + 512) / 1024)
413 #define M_TSO_LEN			66
414 
415 #define MAX_INTS_PER_SEC	8000
416 #define DEFAULT_ITR		(1000000000/(MAX_INTS_PER_SEC * 256))
417 
418 /* Allow common code without TSO */
419 #ifndef CSUM_TSO
420 #define CSUM_TSO	0
421 #endif
422 
423 #define TSO_WORKAROUND	4
424 
425 static SYSCTL_NODE(_hw, OID_AUTO, em, CTLFLAG_RD, 0, "EM driver parameters");
426 
427 static int em_disable_crc_stripping = 0;
428 SYSCTL_INT(_hw_em, OID_AUTO, disable_crc_stripping, CTLFLAG_RDTUN,
429     &em_disable_crc_stripping, 0, "Disable CRC Stripping");
430 
431 static int em_tx_int_delay_dflt = EM_TICKS_TO_USECS(EM_TIDV);
432 static int em_rx_int_delay_dflt = EM_TICKS_TO_USECS(EM_RDTR);
433 SYSCTL_INT(_hw_em, OID_AUTO, tx_int_delay, CTLFLAG_RDTUN, &em_tx_int_delay_dflt,
434     0, "Default transmit interrupt delay in usecs");
435 SYSCTL_INT(_hw_em, OID_AUTO, rx_int_delay, CTLFLAG_RDTUN, &em_rx_int_delay_dflt,
436     0, "Default receive interrupt delay in usecs");
437 
438 static int em_tx_abs_int_delay_dflt = EM_TICKS_TO_USECS(EM_TADV);
439 static int em_rx_abs_int_delay_dflt = EM_TICKS_TO_USECS(EM_RADV);
440 SYSCTL_INT(_hw_em, OID_AUTO, tx_abs_int_delay, CTLFLAG_RDTUN,
441     &em_tx_abs_int_delay_dflt, 0,
442     "Default transmit interrupt delay limit in usecs");
443 SYSCTL_INT(_hw_em, OID_AUTO, rx_abs_int_delay, CTLFLAG_RDTUN,
444     &em_rx_abs_int_delay_dflt, 0,
445     "Default receive interrupt delay limit in usecs");
446 
447 static int em_smart_pwr_down = FALSE;
448 SYSCTL_INT(_hw_em, OID_AUTO, smart_pwr_down, CTLFLAG_RDTUN, &em_smart_pwr_down,
449     0, "Set to true to leave smart power down enabled on newer adapters");
450 
451 /* Controls whether promiscuous also shows bad packets */
452 static int em_debug_sbp = TRUE;
453 SYSCTL_INT(_hw_em, OID_AUTO, sbp, CTLFLAG_RDTUN, &em_debug_sbp, 0,
454     "Show bad packets in promiscuous mode");
455 
456 /* How many packets rxeof tries to clean at a time */
457 static int em_rx_process_limit = 100;
458 SYSCTL_INT(_hw_em, OID_AUTO, rx_process_limit, CTLFLAG_RDTUN,
459     &em_rx_process_limit, 0,
460     "Maximum number of received packets to process "
461     "at a time, -1 means unlimited");
462 
463 /* Energy efficient ethernet - default to OFF */
464 static int eee_setting = 1;
465 SYSCTL_INT(_hw_em, OID_AUTO, eee_setting, CTLFLAG_RDTUN, &eee_setting, 0,
466     "Enable Energy Efficient Ethernet");
467 
468 /*
469 ** Tuneable Interrupt rate
470 */
471 static int em_max_interrupt_rate = 8000;
472 SYSCTL_INT(_hw_em, OID_AUTO, max_interrupt_rate, CTLFLAG_RDTUN,
473     &em_max_interrupt_rate, 0, "Maximum interrupts per second");
474 
475 
476 
477 /* Global used in WOL setup with multiport cards */
478 static int global_quad_port_a = 0;
479 
480 extern struct if_txrx igb_txrx;
481 extern struct if_txrx em_txrx;
482 extern struct if_txrx lem_txrx;
483 
484 static struct if_shared_ctx em_sctx_init = {
485 	.isc_magic = IFLIB_MAGIC,
486 	.isc_q_align = PAGE_SIZE,
487 	.isc_tx_maxsize = EM_TSO_SIZE,
488 	.isc_tx_maxsegsize = PAGE_SIZE,
489 	.isc_rx_maxsize = MJUM9BYTES,
490 	.isc_rx_nsegments = 1,
491 	.isc_rx_maxsegsize = MJUM9BYTES,
492 	.isc_nfl = 1,
493 	.isc_nrxqs = 1,
494 	.isc_ntxqs = 1,
495 	.isc_admin_intrcnt = 1,
496 	.isc_vendor_info = em_vendor_info_array,
497 	.isc_driver_version = em_driver_version,
498 	.isc_driver = &em_if_driver,
499 	.isc_flags = IFLIB_NEED_SCRATCH | IFLIB_TSO_INIT_IP | IFLIB_NEED_ZERO_CSUM,
500 
501 	.isc_nrxd_min = {EM_MIN_RXD},
502 	.isc_ntxd_min = {EM_MIN_TXD},
503 	.isc_nrxd_max = {EM_MAX_RXD},
504 	.isc_ntxd_max = {EM_MAX_TXD},
505 	.isc_nrxd_default = {EM_DEFAULT_RXD},
506 	.isc_ntxd_default = {EM_DEFAULT_TXD},
507 };
508 
509 if_shared_ctx_t em_sctx = &em_sctx_init;
510 
511 
512 static struct if_shared_ctx igb_sctx_init = {
513 	.isc_magic = IFLIB_MAGIC,
514 	.isc_q_align = PAGE_SIZE,
515 	.isc_tx_maxsize = EM_TSO_SIZE,
516 	.isc_tx_maxsegsize = PAGE_SIZE,
517 	.isc_rx_maxsize = MJUM9BYTES,
518 	.isc_rx_nsegments = 1,
519 	.isc_rx_maxsegsize = MJUM9BYTES,
520 	.isc_nfl = 1,
521 	.isc_nrxqs = 1,
522 	.isc_ntxqs = 1,
523 	.isc_admin_intrcnt = 1,
524 	.isc_vendor_info = igb_vendor_info_array,
525 	.isc_driver_version = em_driver_version,
526 	.isc_driver = &em_if_driver,
527 	.isc_flags = IFLIB_NEED_SCRATCH | IFLIB_TSO_INIT_IP | IFLIB_NEED_ZERO_CSUM,
528 
529 	.isc_nrxd_min = {EM_MIN_RXD},
530 	.isc_ntxd_min = {EM_MIN_TXD},
531 	.isc_nrxd_max = {IGB_MAX_RXD},
532 	.isc_ntxd_max = {IGB_MAX_TXD},
533 	.isc_nrxd_default = {EM_DEFAULT_RXD},
534 	.isc_ntxd_default = {EM_DEFAULT_TXD},
535 };
536 
537 if_shared_ctx_t igb_sctx = &igb_sctx_init;
538 
539 /*****************************************************************
540  *
541  * Dump Registers
542  *
543  ****************************************************************/
544 #define IGB_REGS_LEN 739
545 
546 static int em_get_regs(SYSCTL_HANDLER_ARGS)
547 {
548 	struct adapter *adapter = (struct adapter *)arg1;
549 	struct e1000_hw *hw = &adapter->hw;
550 	struct sbuf *sb;
551 	u32 *regs_buff;
552 	int rc;
553 
554 	regs_buff = malloc(sizeof(u32) * IGB_REGS_LEN, M_DEVBUF, M_WAITOK);
555 	memset(regs_buff, 0, IGB_REGS_LEN * sizeof(u32));
556 
557 	rc = sysctl_wire_old_buffer(req, 0);
558 	MPASS(rc == 0);
559 	if (rc != 0) {
560 		free(regs_buff, M_DEVBUF);
561 		return (rc);
562 	}
563 
564 	sb = sbuf_new_for_sysctl(NULL, NULL, 32*400, req);
565 	MPASS(sb != NULL);
566 	if (sb == NULL) {
567 		free(regs_buff, M_DEVBUF);
568 		return (ENOMEM);
569 	}
570 
571 	/* General Registers */
572 	regs_buff[0] = E1000_READ_REG(hw, E1000_CTRL);
573 	regs_buff[1] = E1000_READ_REG(hw, E1000_STATUS);
574 	regs_buff[2] = E1000_READ_REG(hw, E1000_CTRL_EXT);
575 	regs_buff[3] = E1000_READ_REG(hw, E1000_ICR);
576 	regs_buff[4] = E1000_READ_REG(hw, E1000_RCTL);
577 	regs_buff[5] = E1000_READ_REG(hw, E1000_RDLEN(0));
578 	regs_buff[6] = E1000_READ_REG(hw, E1000_RDH(0));
579 	regs_buff[7] = E1000_READ_REG(hw, E1000_RDT(0));
580 	regs_buff[8] = E1000_READ_REG(hw, E1000_RXDCTL(0));
581 	regs_buff[9] = E1000_READ_REG(hw, E1000_RDBAL(0));
582 	regs_buff[10] = E1000_READ_REG(hw, E1000_RDBAH(0));
583 	regs_buff[11] = E1000_READ_REG(hw, E1000_TCTL);
584 	regs_buff[12] = E1000_READ_REG(hw, E1000_TDBAL(0));
585 	regs_buff[13] = E1000_READ_REG(hw, E1000_TDBAH(0));
586 	regs_buff[14] = E1000_READ_REG(hw, E1000_TDLEN(0));
587 	regs_buff[15] = E1000_READ_REG(hw, E1000_TDH(0));
588 	regs_buff[16] = E1000_READ_REG(hw, E1000_TDT(0));
589 	regs_buff[17] = E1000_READ_REG(hw, E1000_TXDCTL(0));
590 	regs_buff[18] = E1000_READ_REG(hw, E1000_TDFH);
591 	regs_buff[19] = E1000_READ_REG(hw, E1000_TDFT);
592 	regs_buff[20] = E1000_READ_REG(hw, E1000_TDFHS);
593 	regs_buff[21] = E1000_READ_REG(hw, E1000_TDFPC);
594 
595 	sbuf_printf(sb, "General Registers\n");
596 	sbuf_printf(sb, "\tCTRL\t %08x\n", regs_buff[0]);
597 	sbuf_printf(sb, "\tSTATUS\t %08x\n", regs_buff[1]);
598 	sbuf_printf(sb, "\tCTRL_EXIT\t %08x\n\n", regs_buff[2]);
599 
600 	sbuf_printf(sb, "Interrupt Registers\n");
601 	sbuf_printf(sb, "\tICR\t %08x\n\n", regs_buff[3]);
602 
603 	sbuf_printf(sb, "RX Registers\n");
604 	sbuf_printf(sb, "\tRCTL\t %08x\n", regs_buff[4]);
605 	sbuf_printf(sb, "\tRDLEN\t %08x\n", regs_buff[5]);
606 	sbuf_printf(sb, "\tRDH\t %08x\n", regs_buff[6]);
607 	sbuf_printf(sb, "\tRDT\t %08x\n", regs_buff[7]);
608 	sbuf_printf(sb, "\tRXDCTL\t %08x\n", regs_buff[8]);
609 	sbuf_printf(sb, "\tRDBAL\t %08x\n", regs_buff[9]);
610 	sbuf_printf(sb, "\tRDBAH\t %08x\n\n", regs_buff[10]);
611 
612 	sbuf_printf(sb, "TX Registers\n");
613 	sbuf_printf(sb, "\tTCTL\t %08x\n", regs_buff[11]);
614 	sbuf_printf(sb, "\tTDBAL\t %08x\n", regs_buff[12]);
615 	sbuf_printf(sb, "\tTDBAH\t %08x\n", regs_buff[13]);
616 	sbuf_printf(sb, "\tTDLEN\t %08x\n", regs_buff[14]);
617 	sbuf_printf(sb, "\tTDH\t %08x\n", regs_buff[15]);
618 	sbuf_printf(sb, "\tTDT\t %08x\n", regs_buff[16]);
619 	sbuf_printf(sb, "\tTXDCTL\t %08x\n", regs_buff[17]);
620 	sbuf_printf(sb, "\tTDFH\t %08x\n", regs_buff[18]);
621 	sbuf_printf(sb, "\tTDFT\t %08x\n", regs_buff[19]);
622 	sbuf_printf(sb, "\tTDFHS\t %08x\n", regs_buff[20]);
623 	sbuf_printf(sb, "\tTDFPC\t %08x\n\n", regs_buff[21]);
624 
625 	free(regs_buff, M_DEVBUF);
626 
627 #ifdef DUMP_DESCS
628 	{
629 		if_softc_ctx_t scctx = adapter->shared;
630 		struct rx_ring *rxr = &rx_que->rxr;
631 		struct tx_ring *txr = &tx_que->txr;
632 		int ntxd = scctx->isc_ntxd[0];
633 		int nrxd = scctx->isc_nrxd[0];
634 		int j;
635 
636 	for (j = 0; j < nrxd; j++) {
637 		u32 staterr = le32toh(rxr->rx_base[j].wb.upper.status_error);
638 		u32 length =  le32toh(rxr->rx_base[j].wb.upper.length);
639 		sbuf_printf(sb, "\tReceive Descriptor Address %d: %08" PRIx64 "  Error:%d  Length:%d\n", j, rxr->rx_base[j].read.buffer_addr, staterr, length);
640 	}
641 
642 	for (j = 0; j < min(ntxd, 256); j++) {
643 		unsigned int *ptr = (unsigned int *)&txr->tx_base[j];
644 
645 		sbuf_printf(sb, "\tTXD[%03d] [0]: %08x [1]: %08x [2]: %08x [3]: %08x  eop: %d DD=%d\n",
646 			    j, ptr[0], ptr[1], ptr[2], ptr[3], buf->eop,
647 			    buf->eop != -1 ? txr->tx_base[buf->eop].upper.fields.status & E1000_TXD_STAT_DD : 0);
648 
649 	}
650 	}
651 #endif
652 
653 	rc = sbuf_finish(sb);
654 	sbuf_delete(sb);
655 	return(rc);
656 }
657 
658 static void *
659 em_register(device_t dev)
660 {
661 	return (em_sctx);
662 }
663 
664 static void *
665 igb_register(device_t dev)
666 {
667 	return (igb_sctx);
668 }
669 
670 static int
671 em_set_num_queues(if_ctx_t ctx)
672 {
673 	struct adapter *adapter = iflib_get_softc(ctx);
674 	int maxqueues;
675 
676 	/* Sanity check based on HW */
677 	switch (adapter->hw.mac.type) {
678 	case e1000_82576:
679 	case e1000_82580:
680 	case e1000_i350:
681 	case e1000_i354:
682 		maxqueues = 8;
683 		break;
684 	case e1000_i210:
685 	case e1000_82575:
686 		maxqueues = 4;
687 		break;
688 	case e1000_i211:
689 	case e1000_82574:
690 		maxqueues = 2;
691 		break;
692 	default:
693 		maxqueues = 1;
694 		break;
695 	}
696 
697 	return (maxqueues);
698 }
699 
700 
701 #define EM_CAPS \
702 	IFCAP_TSO4 | IFCAP_TXCSUM | IFCAP_LRO | IFCAP_RXCSUM | IFCAP_VLAN_HWFILTER | IFCAP_WOL_MAGIC | \
703 	IFCAP_WOL_MCAST | IFCAP_WOL | IFCAP_VLAN_HWTSO | IFCAP_HWCSUM | IFCAP_VLAN_HWTAGGING | \
704 	IFCAP_VLAN_HWCSUM | IFCAP_VLAN_HWTSO | IFCAP_VLAN_MTU;
705 
706 #define IGB_CAPS \
707 	IFCAP_TSO4 | IFCAP_TXCSUM | IFCAP_LRO | IFCAP_RXCSUM | IFCAP_VLAN_HWFILTER | IFCAP_WOL_MAGIC | \
708 	IFCAP_WOL_MCAST | IFCAP_WOL | IFCAP_VLAN_HWTSO | IFCAP_HWCSUM | IFCAP_VLAN_HWTAGGING | IFCAP_VLAN_HWCSUM | \
709 	IFCAP_VLAN_HWTSO | IFCAP_VLAN_MTU | IFCAP_TXCSUM_IPV6 | IFCAP_HWCSUM_IPV6 | IFCAP_JUMBO_MTU;
710 
711 /*********************************************************************
712  *  Device initialization routine
713  *
714  *  The attach entry point is called when the driver is being loaded.
715  *  This routine identifies the type of hardware, allocates all resources
716  *  and initializes the hardware.
717  *
718  *  return 0 on success, positive on failure
719  *********************************************************************/
720 
721 static int
722 em_if_attach_pre(if_ctx_t ctx)
723 {
724 	struct adapter *adapter;
725 	if_softc_ctx_t scctx;
726 	device_t dev;
727 	struct e1000_hw *hw;
728 	int error = 0;
729 
730 	INIT_DEBUGOUT("em_if_attach_pre begin");
731 	dev = iflib_get_dev(ctx);
732 	adapter = iflib_get_softc(ctx);
733 
734 	if (resource_disabled("em", device_get_unit(dev))) {
735 		device_printf(dev, "Disabled by device hint\n");
736 		return (ENXIO);
737 	}
738 
739 	adapter->ctx = 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 	/* Set isc_msix_bar */
777 	scctx->isc_msix_bar = PCIR_BAR(EM_MSIX_BAR);
778 	scctx->isc_tx_nsegments = EM_MAX_SCATTER;
779 	scctx->isc_tx_tso_segments_max = scctx->isc_tx_nsegments;
780 	scctx->isc_tx_tso_size_max = EM_TSO_SIZE;
781 	scctx->isc_tx_tso_segsize_max = EM_TSO_SEG_SIZE;
782 	scctx->isc_nrxqsets_max = scctx->isc_ntxqsets_max = em_set_num_queues(ctx);
783 	device_printf(dev, "attach_pre capping queues at %d\n", scctx->isc_ntxqsets_max);
784 
785 	scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP | CSUM_IP_TSO;
786 
787 
788 	if (adapter->hw.mac.type >= igb_mac_min) {
789 		int try_second_bar;
790 
791 		scctx->isc_txqsizes[0] = roundup2(scctx->isc_ntxd[0] * sizeof(union e1000_adv_tx_desc), EM_DBA_ALIGN);
792 		scctx->isc_rxqsizes[0] = roundup2(scctx->isc_nrxd[0] * sizeof(union e1000_adv_rx_desc), EM_DBA_ALIGN);
793 		scctx->isc_txd_size[0] = sizeof(union e1000_adv_tx_desc);
794 		scctx->isc_rxd_size[0] = sizeof(union e1000_adv_rx_desc);
795 		scctx->isc_txrx = &igb_txrx;
796 		scctx->isc_capenable = IGB_CAPS;
797 		scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP | CSUM_TSO | CSUM_IP6_TCP \
798 			| CSUM_IP6_UDP | CSUM_IP6_TCP;
799 		if (adapter->hw.mac.type != e1000_82575)
800 			scctx->isc_tx_csum_flags |= CSUM_SCTP | CSUM_IP6_SCTP;
801 
802 		/*
803 		** Some new devices, as with ixgbe, now may
804 		** use a different BAR, so we need to keep
805 		** track of which is used.
806 		*/
807 		try_second_bar = pci_read_config(dev, scctx->isc_msix_bar, 4);
808 		if (try_second_bar == 0)
809 			scctx->isc_msix_bar += 4;
810 
811 	} else if (adapter->hw.mac.type >= em_mac_min) {
812 		scctx->isc_txqsizes[0] = roundup2(scctx->isc_ntxd[0]* sizeof(struct e1000_tx_desc), EM_DBA_ALIGN);
813 		scctx->isc_rxqsizes[0] = roundup2(scctx->isc_nrxd[0] * sizeof(union e1000_rx_desc_extended), EM_DBA_ALIGN);
814 		scctx->isc_txd_size[0] = sizeof(struct e1000_tx_desc);
815 		scctx->isc_rxd_size[0] = sizeof(union e1000_rx_desc_extended);
816 		scctx->isc_txrx = &em_txrx;
817 		scctx->isc_capenable = EM_CAPS;
818 		scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP | CSUM_IP_TSO;
819 	} else {
820 		scctx->isc_txqsizes[0] = roundup2((scctx->isc_ntxd[0] + 1) * sizeof(struct e1000_tx_desc), EM_DBA_ALIGN);
821 		scctx->isc_rxqsizes[0] = roundup2((scctx->isc_nrxd[0] + 1) * sizeof(struct e1000_rx_desc), EM_DBA_ALIGN);
822 		scctx->isc_txd_size[0] = sizeof(struct e1000_tx_desc);
823 		scctx->isc_rxd_size[0] = sizeof(struct e1000_rx_desc);
824 		scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP | CSUM_IP_TSO;
825 		scctx->isc_txrx = &lem_txrx;
826 		scctx->isc_capenable = EM_CAPS;
827 		if (adapter->hw.mac.type < e1000_82543)
828 			scctx->isc_capenable &= ~(IFCAP_HWCSUM|IFCAP_VLAN_HWCSUM);
829 		scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP | CSUM_IP_TSO;
830 		scctx->isc_msix_bar = 0;
831 	}
832 
833 	/* Setup PCI resources */
834 	if (em_allocate_pci_resources(ctx)) {
835 		device_printf(dev, "Allocation of PCI resources failed\n");
836 		error = ENXIO;
837 		goto err_pci;
838 	}
839 
840 	/*
841 	** For ICH8 and family we need to
842 	** map the flash memory, and this
843 	** must happen after the MAC is
844 	** identified
845 	*/
846 	if ((hw->mac.type == e1000_ich8lan) ||
847 	    (hw->mac.type == e1000_ich9lan) ||
848 	    (hw->mac.type == e1000_ich10lan) ||
849 	    (hw->mac.type == e1000_pchlan) ||
850 	    (hw->mac.type == e1000_pch2lan) ||
851 	    (hw->mac.type == e1000_pch_lpt)) {
852 		int rid = EM_BAR_TYPE_FLASH;
853 		adapter->flash = bus_alloc_resource_any(dev,
854 		    SYS_RES_MEMORY, &rid, RF_ACTIVE);
855 		if (adapter->flash == NULL) {
856 			device_printf(dev, "Mapping of Flash failed\n");
857 			error = ENXIO;
858 			goto err_pci;
859 		}
860 		/* This is used in the shared code */
861 		hw->flash_address = (u8 *)adapter->flash;
862 		adapter->osdep.flash_bus_space_tag =
863 		    rman_get_bustag(adapter->flash);
864 		adapter->osdep.flash_bus_space_handle =
865 		    rman_get_bushandle(adapter->flash);
866 	}
867 	/*
868 	** In the new SPT device flash is not  a
869 	** separate BAR, rather it is also in BAR0,
870 	** so use the same tag and an offset handle for the
871 	** FLASH read/write macros in the shared code.
872 	*/
873 	else if (hw->mac.type >= e1000_pch_spt) {
874 		adapter->osdep.flash_bus_space_tag =
875 		    adapter->osdep.mem_bus_space_tag;
876 		adapter->osdep.flash_bus_space_handle =
877 		    adapter->osdep.mem_bus_space_handle
878 		    + E1000_FLASH_BASE_ADDR;
879 	}
880 
881 	/* Do Shared Code initialization */
882 	error = e1000_setup_init_funcs(hw, TRUE);
883 	if (error) {
884 		device_printf(dev, "Setup of Shared code failed, error %d\n",
885 		    error);
886 		error = ENXIO;
887 		goto err_pci;
888 	}
889 
890 	em_setup_msix(ctx);
891 	e1000_get_bus_info(hw);
892 
893 	/* Set up some sysctls for the tunable interrupt delays */
894 	em_add_int_delay_sysctl(adapter, "rx_int_delay",
895 	    "receive interrupt delay in usecs", &adapter->rx_int_delay,
896 	    E1000_REGISTER(hw, E1000_RDTR), em_rx_int_delay_dflt);
897 	em_add_int_delay_sysctl(adapter, "tx_int_delay",
898 	    "transmit interrupt delay in usecs", &adapter->tx_int_delay,
899 	    E1000_REGISTER(hw, E1000_TIDV), em_tx_int_delay_dflt);
900 	em_add_int_delay_sysctl(adapter, "rx_abs_int_delay",
901 	    "receive interrupt delay limit in usecs",
902 	    &adapter->rx_abs_int_delay,
903 	    E1000_REGISTER(hw, E1000_RADV),
904 	    em_rx_abs_int_delay_dflt);
905 	em_add_int_delay_sysctl(adapter, "tx_abs_int_delay",
906 	    "transmit interrupt delay limit in usecs",
907 	    &adapter->tx_abs_int_delay,
908 	    E1000_REGISTER(hw, E1000_TADV),
909 	    em_tx_abs_int_delay_dflt);
910 	em_add_int_delay_sysctl(adapter, "itr",
911 	    "interrupt delay limit in usecs/4",
912 	    &adapter->tx_itr,
913 	    E1000_REGISTER(hw, E1000_ITR),
914 	    DEFAULT_ITR);
915 
916 	hw->mac.autoneg = DO_AUTO_NEG;
917 	hw->phy.autoneg_wait_to_complete = FALSE;
918 	hw->phy.autoneg_advertised = AUTONEG_ADV_DEFAULT;
919 
920 	if (adapter->hw.mac.type < em_mac_min) {
921 		e1000_init_script_state_82541(&adapter->hw, TRUE);
922 		e1000_set_tbi_compatibility_82543(&adapter->hw, TRUE);
923 	}
924 	/* Copper options */
925 	if (hw->phy.media_type == e1000_media_type_copper) {
926 		hw->phy.mdix = AUTO_ALL_MODES;
927 		hw->phy.disable_polarity_correction = FALSE;
928 		hw->phy.ms_type = EM_MASTER_SLAVE;
929 	}
930 
931 	/*
932 	 * Set the frame limits assuming
933 	 * standard ethernet sized frames.
934 	 */
935 	scctx->isc_max_frame_size = adapter->hw.mac.max_frame_size =
936 	    ETHERMTU + ETHER_HDR_LEN + ETHERNET_FCS_SIZE;
937 
938 	/*
939 	 * This controls when hardware reports transmit completion
940 	 * status.
941 	 */
942 	hw->mac.report_tx_early = 1;
943 
944 	/* Allocate multicast array memory. */
945 	adapter->mta = malloc(sizeof(u8) * ETH_ADDR_LEN *
946 	    MAX_NUM_MULTICAST_ADDRESSES, M_DEVBUF, M_NOWAIT);
947 	if (adapter->mta == NULL) {
948 		device_printf(dev, "Can not allocate multicast setup array\n");
949 		error = ENOMEM;
950 		goto err_late;
951 	}
952 
953 	/* Check SOL/IDER usage */
954 	if (e1000_check_reset_block(hw))
955 		device_printf(dev, "PHY reset is blocked"
956 			      " due to SOL/IDER session.\n");
957 
958 	/* Sysctl for setting Energy Efficient Ethernet */
959 	hw->dev_spec.ich8lan.eee_disable = eee_setting;
960 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
961 	    SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
962 	    OID_AUTO, "eee_control", CTLTYPE_INT|CTLFLAG_RW,
963 	    adapter, 0, em_sysctl_eee, "I",
964 	    "Disable Energy Efficient Ethernet");
965 
966 	/*
967 	** Start from a known state, this is
968 	** important in reading the nvm and
969 	** mac from that.
970 	*/
971 	e1000_reset_hw(hw);
972 
973 	/* Make sure we have a good EEPROM before we read from it */
974 	if (e1000_validate_nvm_checksum(hw) < 0) {
975 		/*
976 		** Some PCI-E parts fail the first check due to
977 		** the link being in sleep state, call it again,
978 		** if it fails a second time its a real issue.
979 		*/
980 		if (e1000_validate_nvm_checksum(hw) < 0) {
981 			device_printf(dev,
982 			    "The EEPROM Checksum Is Not Valid\n");
983 			error = EIO;
984 			goto err_late;
985 		}
986 	}
987 
988 	/* Copy the permanent MAC address out of the EEPROM */
989 	if (e1000_read_mac_addr(hw) < 0) {
990 		device_printf(dev, "EEPROM read error while reading MAC"
991 			      " address\n");
992 		error = EIO;
993 		goto err_late;
994 	}
995 
996 	if (!em_is_valid_ether_addr(hw->mac.addr)) {
997 		device_printf(dev, "Invalid MAC address\n");
998 		error = EIO;
999 		goto err_late;
1000 	}
1001 
1002 	/* Disable ULP support */
1003 	e1000_disable_ulp_lpt_lp(hw, TRUE);
1004 
1005 	/*
1006 	 * Get Wake-on-Lan and Management info for later use
1007 	 */
1008 	em_get_wakeup(ctx);
1009 
1010 	iflib_set_mac(ctx, hw->mac.addr);
1011 
1012 	return (0);
1013 
1014 err_late:
1015 	em_release_hw_control(adapter);
1016 err_pci:
1017 	em_free_pci_resources(ctx);
1018 	free(adapter->mta, M_DEVBUF);
1019 
1020 	return (error);
1021 }
1022 
1023 static int
1024 em_if_attach_post(if_ctx_t ctx)
1025 {
1026 	struct adapter *adapter = iflib_get_softc(ctx);
1027 	struct e1000_hw *hw = &adapter->hw;
1028 	int error = 0;
1029 
1030 	/* Setup OS specific network interface */
1031 	error = em_setup_interface(ctx);
1032 	if (error != 0) {
1033 		goto err_late;
1034 	}
1035 
1036 	em_reset(ctx);
1037 
1038 	/* Initialize statistics */
1039 	em_update_stats_counters(adapter);
1040 	hw->mac.get_link_status = 1;
1041 	em_if_update_admin_status(ctx);
1042 	em_add_hw_stats(adapter);
1043 
1044 	/* Non-AMT based hardware can now take control from firmware */
1045 	if (adapter->has_manage && !adapter->has_amt)
1046 		em_get_hw_control(adapter);
1047 
1048 	INIT_DEBUGOUT("em_if_attach_post: end");
1049 
1050 	return (error);
1051 
1052 err_late:
1053 	em_release_hw_control(adapter);
1054 	em_free_pci_resources(ctx);
1055 	em_if_queues_free(ctx);
1056 	free(adapter->mta, M_DEVBUF);
1057 
1058 	return (error);
1059 }
1060 
1061 /*********************************************************************
1062  *  Device removal routine
1063  *
1064  *  The detach entry point is called when the driver is being removed.
1065  *  This routine stops the adapter and deallocates all the resources
1066  *  that were allocated for driver operation.
1067  *
1068  *  return 0 on success, positive on failure
1069  *********************************************************************/
1070 
1071 static int
1072 em_if_detach(if_ctx_t ctx)
1073 {
1074 	struct adapter	*adapter = iflib_get_softc(ctx);
1075 
1076 	INIT_DEBUGOUT("em_detach: begin");
1077 
1078 	e1000_phy_hw_reset(&adapter->hw);
1079 
1080 	em_release_manageability(adapter);
1081 	em_release_hw_control(adapter);
1082 	em_free_pci_resources(ctx);
1083 
1084 	return (0);
1085 }
1086 
1087 /*********************************************************************
1088  *
1089  *  Shutdown entry point
1090  *
1091  **********************************************************************/
1092 
1093 static int
1094 em_if_shutdown(if_ctx_t ctx)
1095 {
1096 	return em_if_suspend(ctx);
1097 }
1098 
1099 /*
1100  * Suspend/resume device methods.
1101  */
1102 static int
1103 em_if_suspend(if_ctx_t ctx)
1104 {
1105 	struct adapter *adapter = iflib_get_softc(ctx);
1106 
1107 	em_release_manageability(adapter);
1108 	em_release_hw_control(adapter);
1109 	em_enable_wakeup(ctx);
1110 	return (0);
1111 }
1112 
1113 static int
1114 em_if_resume(if_ctx_t ctx)
1115 {
1116 	struct adapter *adapter = iflib_get_softc(ctx);
1117 
1118 	if (adapter->hw.mac.type == e1000_pch2lan)
1119 		e1000_resume_workarounds_pchlan(&adapter->hw);
1120 	em_if_init(ctx);
1121 	em_init_manageability(adapter);
1122 
1123 	return(0);
1124 }
1125 
1126 static int
1127 em_if_mtu_set(if_ctx_t ctx, uint32_t mtu)
1128 {
1129 	int max_frame_size;
1130 	struct adapter *adapter = iflib_get_softc(ctx);
1131 	if_softc_ctx_t scctx = iflib_get_softc_ctx(ctx);
1132 
1133 	 IOCTL_DEBUGOUT("ioctl rcv'd: SIOCSIFMTU (Set Interface MTU)");
1134 
1135 	switch (adapter->hw.mac.type) {
1136 	case e1000_82571:
1137 	case e1000_82572:
1138 	case e1000_ich9lan:
1139 	case e1000_ich10lan:
1140 	case e1000_pch2lan:
1141 	case e1000_pch_lpt:
1142 	case e1000_pch_spt:
1143 	case e1000_pch_cnp:
1144 	case e1000_82574:
1145 	case e1000_82583:
1146 	case e1000_80003es2lan:
1147 		/* 9K Jumbo Frame size */
1148 		max_frame_size = 9234;
1149 		break;
1150 	case e1000_pchlan:
1151 		max_frame_size = 4096;
1152 		break;
1153 	case e1000_82542:
1154 	case e1000_ich8lan:
1155 		/* Adapters that do not support jumbo frames */
1156 		max_frame_size = ETHER_MAX_LEN;
1157 		break;
1158 	default:
1159 		if (adapter->hw.mac.type >= igb_mac_min)
1160 			max_frame_size = 9234;
1161 		else /* lem */
1162 			max_frame_size = MAX_JUMBO_FRAME_SIZE;
1163 	}
1164 	if (mtu > max_frame_size - ETHER_HDR_LEN - ETHER_CRC_LEN) {
1165 		return (EINVAL);
1166 	}
1167 
1168 	scctx->isc_max_frame_size = adapter->hw.mac.max_frame_size =
1169 	    mtu + ETHER_HDR_LEN + ETHER_CRC_LEN;
1170 	return (0);
1171 }
1172 
1173 /*********************************************************************
1174  *  Init entry point
1175  *
1176  *  This routine is used in two ways. It is used by the stack as
1177  *  init entry point in network interface structure. It is also used
1178  *  by the driver as a hw/sw initialization routine to get to a
1179  *  consistent state.
1180  *
1181  *  return 0 on success, positive on failure
1182  **********************************************************************/
1183 
1184 static void
1185 em_if_init(if_ctx_t ctx)
1186 {
1187 	struct adapter *adapter = iflib_get_softc(ctx);
1188 	struct ifnet *ifp = iflib_get_ifp(ctx);
1189 	struct em_tx_queue *tx_que;
1190 	int i;
1191 	INIT_DEBUGOUT("em_if_init: begin");
1192 
1193 	/* Get the latest mac address, User can use a LAA */
1194 	bcopy(if_getlladdr(ifp), adapter->hw.mac.addr,
1195 	    ETHER_ADDR_LEN);
1196 
1197 	/* Put the address into the Receive Address Array */
1198 	e1000_rar_set(&adapter->hw, adapter->hw.mac.addr, 0);
1199 
1200 	/*
1201 	 * With the 82571 adapter, RAR[0] may be overwritten
1202 	 * when the other port is reset, we make a duplicate
1203 	 * in RAR[14] for that eventuality, this assures
1204 	 * the interface continues to function.
1205 	 */
1206 	if (adapter->hw.mac.type == e1000_82571) {
1207 		e1000_set_laa_state_82571(&adapter->hw, TRUE);
1208 		e1000_rar_set(&adapter->hw, adapter->hw.mac.addr,
1209 		    E1000_RAR_ENTRIES - 1);
1210 	}
1211 
1212 
1213 	/* Initialize the hardware */
1214 	em_reset(ctx);
1215 	em_if_update_admin_status(ctx);
1216 
1217 	for (i = 0, tx_que = adapter->tx_queues; i < adapter->tx_num_queues; i++, tx_que++) {
1218 		struct tx_ring *txr = &tx_que->txr;
1219 
1220 		txr->tx_rs_cidx = txr->tx_rs_pidx = txr->tx_cidx_processed = 0;
1221 	}
1222 
1223 	/* Setup VLAN support, basic and offload if available */
1224 	E1000_WRITE_REG(&adapter->hw, E1000_VET, ETHERTYPE_VLAN);
1225 
1226 	/* Clear bad data from Rx FIFOs */
1227 	if (adapter->hw.mac.type >= igb_mac_min)
1228 		e1000_rx_fifo_flush_82575(&adapter->hw);
1229 
1230 	/* Configure for OS presence */
1231 	em_init_manageability(adapter);
1232 
1233 	/* Prepare transmit descriptors and buffers */
1234 	em_initialize_transmit_unit(ctx);
1235 
1236 	/* Setup Multicast table */
1237 	em_if_multi_set(ctx);
1238 
1239 	/*
1240 	 * Figure out the desired mbuf
1241 	 * pool for doing jumbos
1242 	 */
1243 	if (adapter->hw.mac.max_frame_size <= 2048)
1244 		adapter->rx_mbuf_sz = MCLBYTES;
1245 #ifndef CONTIGMALLOC_WORKS
1246 	else
1247 		adapter->rx_mbuf_sz = MJUMPAGESIZE;
1248 #else
1249 	else if (adapter->hw.mac.max_frame_size <= 4096)
1250 		adapter->rx_mbuf_sz = MJUMPAGESIZE;
1251 	else
1252 		adapter->rx_mbuf_sz = MJUM9BYTES;
1253 #endif
1254 	em_initialize_receive_unit(ctx);
1255 
1256 	/* Use real VLAN Filter support? */
1257 	if (if_getcapenable(ifp) & IFCAP_VLAN_HWTAGGING) {
1258 		if (if_getcapenable(ifp) & IFCAP_VLAN_HWFILTER)
1259 			/* Use real VLAN Filter support */
1260 			em_setup_vlan_hw_support(adapter);
1261 		else {
1262 			u32 ctrl;
1263 			ctrl = E1000_READ_REG(&adapter->hw, E1000_CTRL);
1264 			ctrl |= E1000_CTRL_VME;
1265 			E1000_WRITE_REG(&adapter->hw, E1000_CTRL, ctrl);
1266 		}
1267 	}
1268 
1269 	/* Don't lose promiscuous settings */
1270 	em_if_set_promisc(ctx, IFF_PROMISC);
1271 	e1000_clear_hw_cntrs_base_generic(&adapter->hw);
1272 
1273 	/* MSI/X configuration for 82574 */
1274 	if (adapter->hw.mac.type == e1000_82574) {
1275 		int tmp = E1000_READ_REG(&adapter->hw, E1000_CTRL_EXT);
1276 
1277 		tmp |= E1000_CTRL_EXT_PBA_CLR;
1278 		E1000_WRITE_REG(&adapter->hw, E1000_CTRL_EXT, tmp);
1279 		/* Set the IVAR - interrupt vector routing. */
1280 		E1000_WRITE_REG(&adapter->hw, E1000_IVAR, adapter->ivars);
1281 	} else if (adapter->intr_type == IFLIB_INTR_MSIX) /* Set up queue routing */
1282 		igb_configure_queues(adapter);
1283 
1284 	/* this clears any pending interrupts */
1285 	E1000_READ_REG(&adapter->hw, E1000_ICR);
1286 	E1000_WRITE_REG(&adapter->hw, E1000_ICS, E1000_ICS_LSC);
1287 
1288 	/* AMT based hardware can now take control from firmware */
1289 	if (adapter->has_manage && adapter->has_amt)
1290 		em_get_hw_control(adapter);
1291 
1292 	/* Set Energy Efficient Ethernet */
1293 	if (adapter->hw.mac.type >= igb_mac_min &&
1294 	    adapter->hw.phy.media_type == e1000_media_type_copper) {
1295 		if (adapter->hw.mac.type == e1000_i354)
1296 			e1000_set_eee_i354(&adapter->hw, TRUE, TRUE);
1297 		else
1298 			e1000_set_eee_i350(&adapter->hw, TRUE, TRUE);
1299 	}
1300 }
1301 
1302 /*********************************************************************
1303  *
1304  *  Fast Legacy/MSI Combined Interrupt Service routine
1305  *
1306  *********************************************************************/
1307 int
1308 em_intr(void *arg)
1309 {
1310 	struct adapter *adapter = arg;
1311 	if_ctx_t ctx = adapter->ctx;
1312 	u32 reg_icr;
1313 
1314 	reg_icr = E1000_READ_REG(&adapter->hw, E1000_ICR);
1315 
1316 	if (adapter->intr_type != IFLIB_INTR_LEGACY)
1317 		goto skip_stray;
1318 	/* Hot eject? */
1319 	if (reg_icr == 0xffffffff)
1320 		return FILTER_STRAY;
1321 
1322 	/* Definitely not our interrupt. */
1323 	if (reg_icr == 0x0)
1324 		return FILTER_STRAY;
1325 
1326 	/*
1327 	 * Starting with the 82571 chip, bit 31 should be used to
1328 	 * determine whether the interrupt belongs to us.
1329 	 */
1330 	if (adapter->hw.mac.type >= e1000_82571 &&
1331 	    (reg_icr & E1000_ICR_INT_ASSERTED) == 0)
1332 		return FILTER_STRAY;
1333 
1334 skip_stray:
1335 	/* Link status change */
1336 	if (reg_icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) {
1337 		adapter->hw.mac.get_link_status = 1;
1338 		iflib_admin_intr_deferred(ctx);
1339 	}
1340 
1341 	if (reg_icr & E1000_ICR_RXO)
1342 		adapter->rx_overruns++;
1343 
1344 	return (FILTER_SCHEDULE_THREAD);
1345 }
1346 
1347 static void
1348 igb_rx_enable_queue(struct adapter *adapter, struct em_rx_queue *rxq)
1349 {
1350 	E1000_WRITE_REG(&adapter->hw, E1000_EIMS, rxq->eims);
1351 }
1352 
1353 static void
1354 em_rx_enable_queue(struct adapter *adapter, struct em_rx_queue *rxq)
1355 {
1356 	E1000_WRITE_REG(&adapter->hw, E1000_IMS, rxq->eims);
1357 }
1358 
1359 static void
1360 igb_tx_enable_queue(struct adapter *adapter, struct em_tx_queue *txq)
1361 {
1362 	E1000_WRITE_REG(&adapter->hw, E1000_EIMS, txq->eims);
1363 }
1364 
1365 static void
1366 em_tx_enable_queue(struct adapter *adapter, struct em_tx_queue *txq)
1367 {
1368 	E1000_WRITE_REG(&adapter->hw, E1000_IMS, txq->eims);
1369 }
1370 
1371 static int
1372 em_if_rx_queue_intr_enable(if_ctx_t ctx, uint16_t rxqid)
1373 {
1374 	struct adapter *adapter = iflib_get_softc(ctx);
1375 	struct em_rx_queue *rxq = &adapter->rx_queues[rxqid];
1376 
1377 	if (adapter->hw.mac.type >= igb_mac_min)
1378 		igb_rx_enable_queue(adapter, rxq);
1379 	else
1380 		em_rx_enable_queue(adapter, rxq);
1381 	return (0);
1382 }
1383 
1384 static int
1385 em_if_tx_queue_intr_enable(if_ctx_t ctx, uint16_t txqid)
1386 {
1387 	struct adapter *adapter = iflib_get_softc(ctx);
1388 	struct em_tx_queue *txq = &adapter->tx_queues[txqid];
1389 
1390 	if (adapter->hw.mac.type >= igb_mac_min)
1391 		igb_tx_enable_queue(adapter, txq);
1392 	else
1393 		em_tx_enable_queue(adapter, txq);
1394 	return (0);
1395 }
1396 
1397 /*********************************************************************
1398  *
1399  *  MSIX RX Interrupt Service routine
1400  *
1401  **********************************************************************/
1402 static int
1403 em_msix_que(void *arg)
1404 {
1405 	struct em_rx_queue *que = arg;
1406 
1407 	++que->irqs;
1408 
1409 	return (FILTER_SCHEDULE_THREAD);
1410 }
1411 
1412 /*********************************************************************
1413  *
1414  *  MSIX Link Fast Interrupt Service routine
1415  *
1416  **********************************************************************/
1417 static int
1418 em_msix_link(void *arg)
1419 {
1420 	struct adapter *adapter = arg;
1421 	u32 reg_icr;
1422 
1423 	++adapter->link_irq;
1424 	MPASS(adapter->hw.back != NULL);
1425 	reg_icr = E1000_READ_REG(&adapter->hw, E1000_ICR);
1426 
1427 	if (reg_icr & E1000_ICR_RXO)
1428 		adapter->rx_overruns++;
1429 
1430 	if (reg_icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) {
1431 		em_handle_link(adapter->ctx);
1432 	} else {
1433 		E1000_WRITE_REG(&adapter->hw, E1000_IMS,
1434 				EM_MSIX_LINK | E1000_IMS_LSC);
1435 		if (adapter->hw.mac.type >= igb_mac_min)
1436 			E1000_WRITE_REG(&adapter->hw, E1000_EIMS, adapter->link_mask);
1437 	}
1438 
1439 	/*
1440 	 * Because we must read the ICR for this interrupt
1441 	 * it may clear other causes using autoclear, for
1442 	 * this reason we simply create a soft interrupt
1443 	 * for all these vectors.
1444 	 */
1445 	if (reg_icr && adapter->hw.mac.type < igb_mac_min) {
1446 		E1000_WRITE_REG(&adapter->hw,
1447 			E1000_ICS, adapter->ims);
1448 	}
1449 
1450 	return (FILTER_HANDLED);
1451 }
1452 
1453 static void
1454 em_handle_link(void *context)
1455 {
1456 	if_ctx_t ctx = context;
1457 	struct adapter *adapter = iflib_get_softc(ctx);
1458 
1459 	adapter->hw.mac.get_link_status = 1;
1460 	iflib_admin_intr_deferred(ctx);
1461 }
1462 
1463 
1464 /*********************************************************************
1465  *
1466  *  Media Ioctl callback
1467  *
1468  *  This routine is called whenever the user queries the status of
1469  *  the interface using ifconfig.
1470  *
1471  **********************************************************************/
1472 static void
1473 em_if_media_status(if_ctx_t ctx, struct ifmediareq *ifmr)
1474 {
1475 	struct adapter *adapter = iflib_get_softc(ctx);
1476 	u_char fiber_type = IFM_1000_SX;
1477 
1478 	INIT_DEBUGOUT("em_if_media_status: begin");
1479 
1480 	iflib_admin_intr_deferred(ctx);
1481 
1482 	ifmr->ifm_status = IFM_AVALID;
1483 	ifmr->ifm_active = IFM_ETHER;
1484 
1485 	if (!adapter->link_active) {
1486 		return;
1487 	}
1488 
1489 	ifmr->ifm_status |= IFM_ACTIVE;
1490 
1491 	if ((adapter->hw.phy.media_type == e1000_media_type_fiber) ||
1492 	    (adapter->hw.phy.media_type == e1000_media_type_internal_serdes)) {
1493 		if (adapter->hw.mac.type == e1000_82545)
1494 			fiber_type = IFM_1000_LX;
1495 		ifmr->ifm_active |= fiber_type | IFM_FDX;
1496 	} else {
1497 		switch (adapter->link_speed) {
1498 		case 10:
1499 			ifmr->ifm_active |= IFM_10_T;
1500 			break;
1501 		case 100:
1502 			ifmr->ifm_active |= IFM_100_TX;
1503 			break;
1504 		case 1000:
1505 			ifmr->ifm_active |= IFM_1000_T;
1506 			break;
1507 		}
1508 		if (adapter->link_duplex == FULL_DUPLEX)
1509 			ifmr->ifm_active |= IFM_FDX;
1510 		else
1511 			ifmr->ifm_active |= IFM_HDX;
1512 	}
1513 }
1514 
1515 /*********************************************************************
1516  *
1517  *  Media Ioctl callback
1518  *
1519  *  This routine is called when the user changes speed/duplex using
1520  *  media/mediopt option with ifconfig.
1521  *
1522  **********************************************************************/
1523 static int
1524 em_if_media_change(if_ctx_t ctx)
1525 {
1526 	struct adapter *adapter = iflib_get_softc(ctx);
1527 	struct ifmedia *ifm = iflib_get_media(ctx);
1528 
1529 	INIT_DEBUGOUT("em_if_media_change: begin");
1530 
1531 	if (IFM_TYPE(ifm->ifm_media) != IFM_ETHER)
1532 		return (EINVAL);
1533 
1534 	switch (IFM_SUBTYPE(ifm->ifm_media)) {
1535 	case IFM_AUTO:
1536 		adapter->hw.mac.autoneg = DO_AUTO_NEG;
1537 		adapter->hw.phy.autoneg_advertised = AUTONEG_ADV_DEFAULT;
1538 		break;
1539 	case IFM_1000_LX:
1540 	case IFM_1000_SX:
1541 	case IFM_1000_T:
1542 		adapter->hw.mac.autoneg = DO_AUTO_NEG;
1543 		adapter->hw.phy.autoneg_advertised = ADVERTISE_1000_FULL;
1544 		break;
1545 	case IFM_100_TX:
1546 		adapter->hw.mac.autoneg = FALSE;
1547 		adapter->hw.phy.autoneg_advertised = 0;
1548 		if ((ifm->ifm_media & IFM_GMASK) == IFM_FDX)
1549 			adapter->hw.mac.forced_speed_duplex = ADVERTISE_100_FULL;
1550 		else
1551 			adapter->hw.mac.forced_speed_duplex = ADVERTISE_100_HALF;
1552 		break;
1553 	case IFM_10_T:
1554 		adapter->hw.mac.autoneg = FALSE;
1555 		adapter->hw.phy.autoneg_advertised = 0;
1556 		if ((ifm->ifm_media & IFM_GMASK) == IFM_FDX)
1557 			adapter->hw.mac.forced_speed_duplex = ADVERTISE_10_FULL;
1558 		else
1559 			adapter->hw.mac.forced_speed_duplex = ADVERTISE_10_HALF;
1560 		break;
1561 	default:
1562 		device_printf(adapter->dev, "Unsupported media type\n");
1563 	}
1564 
1565 	em_if_init(ctx);
1566 
1567 	return (0);
1568 }
1569 
1570 static int
1571 em_if_set_promisc(if_ctx_t ctx, int flags)
1572 {
1573 	struct adapter *adapter = iflib_get_softc(ctx);
1574 	u32 reg_rctl;
1575 
1576 	em_disable_promisc(ctx);
1577 
1578 	reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
1579 
1580 	if (flags & IFF_PROMISC) {
1581 		reg_rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
1582 		/* Turn this on if you want to see bad packets */
1583 		if (em_debug_sbp)
1584 			reg_rctl |= E1000_RCTL_SBP;
1585 		E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1586 	} else if (flags & IFF_ALLMULTI) {
1587 		reg_rctl |= E1000_RCTL_MPE;
1588 		reg_rctl &= ~E1000_RCTL_UPE;
1589 		E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1590 	}
1591 	return (0);
1592 }
1593 
1594 static void
1595 em_disable_promisc(if_ctx_t ctx)
1596 {
1597 	struct adapter *adapter = iflib_get_softc(ctx);
1598 	struct ifnet *ifp = iflib_get_ifp(ctx);
1599 	u32 reg_rctl;
1600 	int mcnt = 0;
1601 
1602 	reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
1603 	reg_rctl &= (~E1000_RCTL_UPE);
1604 	if (if_getflags(ifp) & IFF_ALLMULTI)
1605 		mcnt = MAX_NUM_MULTICAST_ADDRESSES;
1606 	else
1607 		mcnt = if_multiaddr_count(ifp, MAX_NUM_MULTICAST_ADDRESSES);
1608 	/* Don't disable if in MAX groups */
1609 	if (mcnt < MAX_NUM_MULTICAST_ADDRESSES)
1610 		reg_rctl &=  (~E1000_RCTL_MPE);
1611 	reg_rctl &=  (~E1000_RCTL_SBP);
1612 	E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1613 }
1614 
1615 
1616 /*********************************************************************
1617  *  Multicast Update
1618  *
1619  *  This routine is called whenever multicast address list is updated.
1620  *
1621  **********************************************************************/
1622 
1623 static void
1624 em_if_multi_set(if_ctx_t ctx)
1625 {
1626 	struct adapter *adapter = iflib_get_softc(ctx);
1627 	struct ifnet *ifp = iflib_get_ifp(ctx);
1628 	u32 reg_rctl = 0;
1629 	u8  *mta; /* Multicast array memory */
1630 	int mcnt = 0;
1631 
1632 	IOCTL_DEBUGOUT("em_set_multi: begin");
1633 
1634 	mta = adapter->mta;
1635 	bzero(mta, sizeof(u8) * ETH_ADDR_LEN * MAX_NUM_MULTICAST_ADDRESSES);
1636 
1637 	if (adapter->hw.mac.type == e1000_82542 &&
1638 	    adapter->hw.revision_id == E1000_REVISION_2) {
1639 		reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
1640 		if (adapter->hw.bus.pci_cmd_word & CMD_MEM_WRT_INVALIDATE)
1641 			e1000_pci_clear_mwi(&adapter->hw);
1642 		reg_rctl |= E1000_RCTL_RST;
1643 		E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1644 		msec_delay(5);
1645 	}
1646 
1647 	if_multiaddr_array(ifp, mta, &mcnt, MAX_NUM_MULTICAST_ADDRESSES);
1648 
1649 	if (mcnt >= MAX_NUM_MULTICAST_ADDRESSES) {
1650 		reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
1651 		reg_rctl |= E1000_RCTL_MPE;
1652 		E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1653 	} else
1654 		e1000_update_mc_addr_list(&adapter->hw, mta, mcnt);
1655 
1656 	if (adapter->hw.mac.type == e1000_82542 &&
1657 	    adapter->hw.revision_id == E1000_REVISION_2) {
1658 		reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
1659 		reg_rctl &= ~E1000_RCTL_RST;
1660 		E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1661 		msec_delay(5);
1662 		if (adapter->hw.bus.pci_cmd_word & CMD_MEM_WRT_INVALIDATE)
1663 			e1000_pci_set_mwi(&adapter->hw);
1664 	}
1665 }
1666 
1667 
1668 /*********************************************************************
1669  *  Timer routine
1670  *
1671  *  This routine checks for link status and updates statistics.
1672  *
1673  **********************************************************************/
1674 
1675 static void
1676 em_if_timer(if_ctx_t ctx, uint16_t qid)
1677 {
1678 	struct adapter *adapter = iflib_get_softc(ctx);
1679 	struct em_rx_queue *que;
1680 	int i;
1681 	int trigger = 0;
1682 
1683 	if (qid != 0)
1684 		return;
1685 
1686 	iflib_admin_intr_deferred(ctx);
1687 
1688 	/* Mask to use in the irq trigger */
1689 	if (adapter->intr_type == IFLIB_INTR_MSIX) {
1690 		for (i = 0, que = adapter->rx_queues; i < adapter->rx_num_queues; i++, que++)
1691 			trigger |= que->eims;
1692 	} else {
1693 		trigger = E1000_ICS_RXDMT0;
1694 	}
1695 }
1696 
1697 
1698 static void
1699 em_if_update_admin_status(if_ctx_t ctx)
1700 {
1701 	struct adapter *adapter = iflib_get_softc(ctx);
1702 	struct e1000_hw *hw = &adapter->hw;
1703 	struct ifnet *ifp = iflib_get_ifp(ctx);
1704 	device_t dev = iflib_get_dev(ctx);
1705 	u32 link_check, thstat, ctrl;
1706 
1707 	link_check = thstat = ctrl = 0;
1708 	/* Get the cached link value or read phy for real */
1709 	switch (hw->phy.media_type) {
1710 	case e1000_media_type_copper:
1711 		if (hw->mac.get_link_status) {
1712 			if (hw->mac.type == e1000_pch_spt)
1713 				msec_delay(50);
1714 			/* Do the work to read phy */
1715 			e1000_check_for_link(hw);
1716 			link_check = !hw->mac.get_link_status;
1717 			if (link_check) /* ESB2 fix */
1718 				e1000_cfg_on_link_up(hw);
1719 		} else {
1720 			link_check = TRUE;
1721 		}
1722 		break;
1723 	case e1000_media_type_fiber:
1724 		e1000_check_for_link(hw);
1725 		link_check = (E1000_READ_REG(hw, E1000_STATUS) &
1726 			    E1000_STATUS_LU);
1727 		break;
1728 	case e1000_media_type_internal_serdes:
1729 		e1000_check_for_link(hw);
1730 		link_check = adapter->hw.mac.serdes_has_link;
1731 		break;
1732 	/* VF device is type_unknown */
1733 	case e1000_media_type_unknown:
1734 		e1000_check_for_link(hw);
1735 		link_check = !hw->mac.get_link_status;
1736 		/* FALLTHROUGH */
1737 	default:
1738 		break;
1739 	}
1740 
1741 	/* Check for thermal downshift or shutdown */
1742 	if (hw->mac.type == e1000_i350) {
1743 		thstat = E1000_READ_REG(hw, E1000_THSTAT);
1744 		ctrl = E1000_READ_REG(hw, E1000_CTRL_EXT);
1745 	}
1746 
1747 	/* Now check for a transition */
1748 	if (link_check && (adapter->link_active == 0)) {
1749 		e1000_get_speed_and_duplex(hw, &adapter->link_speed,
1750 		    &adapter->link_duplex);
1751 		/* Check if we must disable SPEED_MODE bit on PCI-E */
1752 		if ((adapter->link_speed != SPEED_1000) &&
1753 		    ((hw->mac.type == e1000_82571) ||
1754 		    (hw->mac.type == e1000_82572))) {
1755 			int tarc0;
1756 			tarc0 = E1000_READ_REG(hw, E1000_TARC(0));
1757 			tarc0 &= ~TARC_SPEED_MODE_BIT;
1758 			E1000_WRITE_REG(hw, E1000_TARC(0), tarc0);
1759 		}
1760 		if (bootverbose)
1761 			device_printf(dev, "Link is up %d Mbps %s\n",
1762 			    adapter->link_speed,
1763 			    ((adapter->link_duplex == FULL_DUPLEX) ?
1764 			    "Full Duplex" : "Half Duplex"));
1765 		adapter->link_active = 1;
1766 		adapter->smartspeed = 0;
1767 		if_setbaudrate(ifp, adapter->link_speed * 1000000);
1768 		if ((ctrl & E1000_CTRL_EXT_LINK_MODE_GMII) &&
1769 		    (thstat & E1000_THSTAT_LINK_THROTTLE))
1770 			device_printf(dev, "Link: thermal downshift\n");
1771 		/* Delay Link Up for Phy update */
1772 		if (((hw->mac.type == e1000_i210) ||
1773 		    (hw->mac.type == e1000_i211)) &&
1774 		    (hw->phy.id == I210_I_PHY_ID))
1775 			msec_delay(I210_LINK_DELAY);
1776 		/* Reset if the media type changed. */
1777 		if ((hw->dev_spec._82575.media_changed) &&
1778 			(adapter->hw.mac.type >= igb_mac_min)) {
1779 			hw->dev_spec._82575.media_changed = false;
1780 			adapter->flags |= IGB_MEDIA_RESET;
1781 			em_reset(ctx);
1782 		}
1783 		iflib_link_state_change(ctx, LINK_STATE_UP, ifp->if_baudrate);
1784 		printf("Link state changed to up\n");
1785 	} else if (!link_check && (adapter->link_active == 1)) {
1786 		if_setbaudrate(ifp, 0);
1787 		adapter->link_speed = 0;
1788 		adapter->link_duplex = 0;
1789 		if (bootverbose)
1790 			device_printf(dev, "Link is Down\n");
1791 		adapter->link_active = 0;
1792 		iflib_link_state_change(ctx, LINK_STATE_DOWN, ifp->if_baudrate);
1793 		printf("link state changed to down\n");
1794 	}
1795 	em_update_stats_counters(adapter);
1796 
1797 	/* Reset LAA into RAR[0] on 82571 */
1798 	if ((adapter->hw.mac.type == e1000_82571) &&
1799 	    e1000_get_laa_state_82571(&adapter->hw))
1800 		e1000_rar_set(&adapter->hw, adapter->hw.mac.addr, 0);
1801 
1802 	if (adapter->hw.mac.type < em_mac_min)
1803 		lem_smartspeed(adapter);
1804 
1805 	E1000_WRITE_REG(&adapter->hw, E1000_IMS, EM_MSIX_LINK | E1000_IMS_LSC);
1806 }
1807 
1808 /*********************************************************************
1809  *
1810  *  This routine disables all traffic on the adapter by issuing a
1811  *  global reset on the MAC and deallocates TX/RX buffers.
1812  *
1813  *  This routine should always be called with BOTH the CORE
1814  *  and TX locks.
1815  **********************************************************************/
1816 
1817 static void
1818 em_if_stop(if_ctx_t ctx)
1819 {
1820 	struct adapter *adapter = iflib_get_softc(ctx);
1821 
1822 	INIT_DEBUGOUT("em_stop: begin");
1823 
1824 	e1000_reset_hw(&adapter->hw);
1825 	if (adapter->hw.mac.type >= e1000_82544)
1826 		E1000_WRITE_REG(&adapter->hw, E1000_WUFC, 0);
1827 
1828 	e1000_led_off(&adapter->hw);
1829 	e1000_cleanup_led(&adapter->hw);
1830 }
1831 
1832 
1833 /*********************************************************************
1834  *
1835  *  Determine hardware revision.
1836  *
1837  **********************************************************************/
1838 static void
1839 em_identify_hardware(if_ctx_t ctx)
1840 {
1841 	device_t dev = iflib_get_dev(ctx);
1842 	struct adapter *adapter = iflib_get_softc(ctx);
1843 
1844 	/* Make sure our PCI config space has the necessary stuff set */
1845 	adapter->hw.bus.pci_cmd_word = pci_read_config(dev, PCIR_COMMAND, 2);
1846 
1847 	/* Save off the information about this board */
1848 	adapter->hw.vendor_id = pci_get_vendor(dev);
1849 	adapter->hw.device_id = pci_get_device(dev);
1850 	adapter->hw.revision_id = pci_read_config(dev, PCIR_REVID, 1);
1851 	adapter->hw.subsystem_vendor_id =
1852 	    pci_read_config(dev, PCIR_SUBVEND_0, 2);
1853 	adapter->hw.subsystem_device_id =
1854 	    pci_read_config(dev, PCIR_SUBDEV_0, 2);
1855 
1856 	/* Do Shared Code Init and Setup */
1857 	if (e1000_set_mac_type(&adapter->hw)) {
1858 		device_printf(dev, "Setup init failure\n");
1859 		return;
1860 	}
1861 }
1862 
1863 static int
1864 em_allocate_pci_resources(if_ctx_t ctx)
1865 {
1866 	struct adapter *adapter = iflib_get_softc(ctx);
1867 	device_t dev = iflib_get_dev(ctx);
1868 	int rid, val;
1869 
1870 	rid = PCIR_BAR(0);
1871 	adapter->memory = bus_alloc_resource_any(dev, SYS_RES_MEMORY,
1872 	    &rid, RF_ACTIVE);
1873 	if (adapter->memory == NULL) {
1874 		device_printf(dev, "Unable to allocate bus resource: memory\n");
1875 		return (ENXIO);
1876 	}
1877 	adapter->osdep.mem_bus_space_tag = rman_get_bustag(adapter->memory);
1878 	adapter->osdep.mem_bus_space_handle =
1879 	    rman_get_bushandle(adapter->memory);
1880 	adapter->hw.hw_addr = (u8 *)&adapter->osdep.mem_bus_space_handle;
1881 
1882 	/* Only older adapters use IO mapping */
1883 	if (adapter->hw.mac.type < em_mac_min &&
1884 	    adapter->hw.mac.type > e1000_82543) {
1885 		/* Figure our where our IO BAR is ? */
1886 		for (rid = PCIR_BAR(0); rid < PCIR_CIS;) {
1887 			val = pci_read_config(dev, rid, 4);
1888 			if (EM_BAR_TYPE(val) == EM_BAR_TYPE_IO) {
1889 				adapter->io_rid = rid;
1890 				break;
1891 			}
1892 			rid += 4;
1893 			/* check for 64bit BAR */
1894 			if (EM_BAR_MEM_TYPE(val) == EM_BAR_MEM_TYPE_64BIT)
1895 				rid += 4;
1896 		}
1897 		if (rid >= PCIR_CIS) {
1898 			device_printf(dev, "Unable to locate IO BAR\n");
1899 			return (ENXIO);
1900 		}
1901 		adapter->ioport = bus_alloc_resource_any(dev,
1902 		    SYS_RES_IOPORT, &adapter->io_rid, RF_ACTIVE);
1903 		if (adapter->ioport == NULL) {
1904 			device_printf(dev, "Unable to allocate bus resource: "
1905 			    "ioport\n");
1906 			return (ENXIO);
1907 		}
1908 		adapter->hw.io_base = 0;
1909 		adapter->osdep.io_bus_space_tag =
1910 		    rman_get_bustag(adapter->ioport);
1911 		adapter->osdep.io_bus_space_handle =
1912 		    rman_get_bushandle(adapter->ioport);
1913 	}
1914 
1915 	adapter->hw.back = &adapter->osdep;
1916 
1917 	return (0);
1918 }
1919 
1920 /*********************************************************************
1921  *
1922  *  Setup the MSIX Interrupt handlers
1923  *
1924  **********************************************************************/
1925 static int
1926 em_if_msix_intr_assign(if_ctx_t ctx, int msix)
1927 {
1928 	struct adapter *adapter = iflib_get_softc(ctx);
1929 	struct em_rx_queue *rx_que = adapter->rx_queues;
1930 	struct em_tx_queue *tx_que = adapter->tx_queues;
1931 	int error, rid, i, vector = 0, rx_vectors;
1932 	char buf[16];
1933 
1934 	/* First set up ring resources */
1935 	for (i = 0; i < adapter->rx_num_queues; i++, rx_que++, vector++) {
1936 		rid = vector + 1;
1937 		snprintf(buf, sizeof(buf), "rxq%d", i);
1938 		error = iflib_irq_alloc_generic(ctx, &rx_que->que_irq, rid, IFLIB_INTR_RXTX, em_msix_que, rx_que, rx_que->me, buf);
1939 		if (error) {
1940 			device_printf(iflib_get_dev(ctx), "Failed to allocate que int %d err: %d", i, error);
1941 			adapter->rx_num_queues = i + 1;
1942 			goto fail;
1943 		}
1944 
1945 		rx_que->msix =  vector;
1946 
1947 		/*
1948 		 * Set the bit to enable interrupt
1949 		 * in E1000_IMS -- bits 20 and 21
1950 		 * are for RX0 and RX1, note this has
1951 		 * NOTHING to do with the MSIX vector
1952 		 */
1953 		if (adapter->hw.mac.type == e1000_82574) {
1954 			rx_que->eims = 1 << (20 + i);
1955 			adapter->ims |= rx_que->eims;
1956 			adapter->ivars |= (8 | rx_que->msix) << (i * 4);
1957 		} else if (adapter->hw.mac.type == e1000_82575)
1958 			rx_que->eims = E1000_EICR_TX_QUEUE0 << vector;
1959 		else
1960 			rx_que->eims = 1 << vector;
1961 	}
1962 	rx_vectors = vector;
1963 
1964 	vector = 0;
1965 	for (i = 0; i < adapter->tx_num_queues; i++, tx_que++, vector++) {
1966 		rid = vector + 1;
1967 		snprintf(buf, sizeof(buf), "txq%d", i);
1968 		tx_que = &adapter->tx_queues[i];
1969 		iflib_softirq_alloc_generic(ctx,
1970 		    &adapter->rx_queues[i % adapter->rx_num_queues].que_irq,
1971 		    IFLIB_INTR_TX, tx_que, tx_que->me, buf);
1972 
1973 		tx_que->msix = (vector % adapter->tx_num_queues);
1974 
1975 		/*
1976 		 * Set the bit to enable interrupt
1977 		 * in E1000_IMS -- bits 22 and 23
1978 		 * are for TX0 and TX1, note this has
1979 		 * NOTHING to do with the MSIX vector
1980 		 */
1981 		if (adapter->hw.mac.type == e1000_82574) {
1982 			tx_que->eims = 1 << (22 + i);
1983 			adapter->ims |= tx_que->eims;
1984 			adapter->ivars |= (8 | tx_que->msix) << (8 + (i * 4));
1985 		} else if (adapter->hw.mac.type == e1000_82575) {
1986 			tx_que->eims = E1000_EICR_TX_QUEUE0 << (i %  adapter->tx_num_queues);
1987 		} else {
1988 			tx_que->eims = 1 << (i %  adapter->tx_num_queues);
1989 		}
1990 	}
1991 
1992 	/* Link interrupt */
1993 	rid = rx_vectors + 1;
1994 	error = iflib_irq_alloc_generic(ctx, &adapter->irq, rid, IFLIB_INTR_ADMIN, em_msix_link, adapter, 0, "aq");
1995 
1996 	if (error) {
1997 		device_printf(iflib_get_dev(ctx), "Failed to register admin handler");
1998 		goto fail;
1999 	}
2000 	adapter->linkvec = rx_vectors;
2001 	if (adapter->hw.mac.type < igb_mac_min) {
2002 		adapter->ivars |=  (8 | rx_vectors) << 16;
2003 		adapter->ivars |= 0x80000000;
2004 	}
2005 	return (0);
2006 fail:
2007 	iflib_irq_free(ctx, &adapter->irq);
2008 	rx_que = adapter->rx_queues;
2009 	for (int i = 0; i < adapter->rx_num_queues; i++, rx_que++)
2010 		iflib_irq_free(ctx, &rx_que->que_irq);
2011 	return (error);
2012 }
2013 
2014 static void
2015 igb_configure_queues(struct adapter *adapter)
2016 {
2017 	struct e1000_hw *hw = &adapter->hw;
2018 	struct em_rx_queue *rx_que;
2019 	struct em_tx_queue *tx_que;
2020 	u32 tmp, ivar = 0, newitr = 0;
2021 
2022 	/* First turn on RSS capability */
2023 	if (adapter->hw.mac.type != e1000_82575)
2024 		E1000_WRITE_REG(hw, E1000_GPIE,
2025 		    E1000_GPIE_MSIX_MODE | E1000_GPIE_EIAME |
2026 		    E1000_GPIE_PBA | E1000_GPIE_NSICR);
2027 
2028 	/* Turn on MSIX */
2029 	switch (adapter->hw.mac.type) {
2030 	case e1000_82580:
2031 	case e1000_i350:
2032 	case e1000_i354:
2033 	case e1000_i210:
2034 	case e1000_i211:
2035 	case e1000_vfadapt:
2036 	case e1000_vfadapt_i350:
2037 		/* RX entries */
2038 		for (int i = 0; i < adapter->rx_num_queues; i++) {
2039 			u32 index = i >> 1;
2040 			ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2041 			rx_que = &adapter->rx_queues[i];
2042 			if (i & 1) {
2043 				ivar &= 0xFF00FFFF;
2044 				ivar |= (rx_que->msix | E1000_IVAR_VALID) << 16;
2045 			} else {
2046 				ivar &= 0xFFFFFF00;
2047 				ivar |= rx_que->msix | E1000_IVAR_VALID;
2048 			}
2049 			E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2050 		}
2051 		/* TX entries */
2052 		for (int i = 0; i < adapter->tx_num_queues; i++) {
2053 			u32 index = i >> 1;
2054 			ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2055 			tx_que = &adapter->tx_queues[i];
2056 			if (i & 1) {
2057 				ivar &= 0x00FFFFFF;
2058 				ivar |= (tx_que->msix | E1000_IVAR_VALID) << 24;
2059 			} else {
2060 				ivar &= 0xFFFF00FF;
2061 				ivar |= (tx_que->msix | E1000_IVAR_VALID) << 8;
2062 			}
2063 			E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2064 			adapter->que_mask |= tx_que->eims;
2065 		}
2066 
2067 		/* And for the link interrupt */
2068 		ivar = (adapter->linkvec | E1000_IVAR_VALID) << 8;
2069 		adapter->link_mask = 1 << adapter->linkvec;
2070 		E1000_WRITE_REG(hw, E1000_IVAR_MISC, ivar);
2071 		break;
2072 	case e1000_82576:
2073 		/* RX entries */
2074 		for (int i = 0; i < adapter->rx_num_queues; i++) {
2075 			u32 index = i & 0x7; /* Each IVAR has two entries */
2076 			ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2077 			rx_que = &adapter->rx_queues[i];
2078 			if (i < 8) {
2079 				ivar &= 0xFFFFFF00;
2080 				ivar |= rx_que->msix | E1000_IVAR_VALID;
2081 			} else {
2082 				ivar &= 0xFF00FFFF;
2083 				ivar |= (rx_que->msix | E1000_IVAR_VALID) << 16;
2084 			}
2085 			E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2086 			adapter->que_mask |= rx_que->eims;
2087 		}
2088 		/* TX entries */
2089 		for (int i = 0; i < adapter->tx_num_queues; i++) {
2090 			u32 index = i & 0x7; /* Each IVAR has two entries */
2091 			ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2092 			tx_que = &adapter->tx_queues[i];
2093 			if (i < 8) {
2094 				ivar &= 0xFFFF00FF;
2095 				ivar |= (tx_que->msix | E1000_IVAR_VALID) << 8;
2096 			} else {
2097 				ivar &= 0x00FFFFFF;
2098 				ivar |= (tx_que->msix | E1000_IVAR_VALID) << 24;
2099 			}
2100 			E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2101 			adapter->que_mask |= tx_que->eims;
2102 		}
2103 
2104 		/* And for the link interrupt */
2105 		ivar = (adapter->linkvec | E1000_IVAR_VALID) << 8;
2106 		adapter->link_mask = 1 << adapter->linkvec;
2107 		E1000_WRITE_REG(hw, E1000_IVAR_MISC, ivar);
2108 		break;
2109 
2110 	case e1000_82575:
2111 		/* enable MSI-X support*/
2112 		tmp = E1000_READ_REG(hw, E1000_CTRL_EXT);
2113 		tmp |= E1000_CTRL_EXT_PBA_CLR;
2114 		/* Auto-Mask interrupts upon ICR read. */
2115 		tmp |= E1000_CTRL_EXT_EIAME;
2116 		tmp |= E1000_CTRL_EXT_IRCA;
2117 		E1000_WRITE_REG(hw, E1000_CTRL_EXT, tmp);
2118 
2119 		/* Queues */
2120 		for (int i = 0; i < adapter->rx_num_queues; i++) {
2121 			rx_que = &adapter->rx_queues[i];
2122 			tmp = E1000_EICR_RX_QUEUE0 << i;
2123 			tmp |= E1000_EICR_TX_QUEUE0 << i;
2124 			rx_que->eims = tmp;
2125 			E1000_WRITE_REG_ARRAY(hw, E1000_MSIXBM(0),
2126 			    i, rx_que->eims);
2127 			adapter->que_mask |= rx_que->eims;
2128 		}
2129 
2130 		/* Link */
2131 		E1000_WRITE_REG(hw, E1000_MSIXBM(adapter->linkvec),
2132 		    E1000_EIMS_OTHER);
2133 		adapter->link_mask |= E1000_EIMS_OTHER;
2134 	default:
2135 		break;
2136 	}
2137 
2138 	/* Set the starting interrupt rate */
2139 	if (em_max_interrupt_rate > 0)
2140 		newitr = (4000000 / em_max_interrupt_rate) & 0x7FFC;
2141 
2142 	if (hw->mac.type == e1000_82575)
2143 		newitr |= newitr << 16;
2144 	else
2145 		newitr |= E1000_EITR_CNT_IGNR;
2146 
2147 	for (int i = 0; i < adapter->rx_num_queues; i++) {
2148 		rx_que = &adapter->rx_queues[i];
2149 		E1000_WRITE_REG(hw, E1000_EITR(rx_que->msix), newitr);
2150 	}
2151 
2152 	return;
2153 }
2154 
2155 static void
2156 em_free_pci_resources(if_ctx_t ctx)
2157 {
2158 	struct adapter *adapter = iflib_get_softc(ctx);
2159 	struct em_rx_queue *que = adapter->rx_queues;
2160 	device_t dev = iflib_get_dev(ctx);
2161 
2162 	/* Release all msix queue resources */
2163 	if (adapter->intr_type == IFLIB_INTR_MSIX)
2164 		iflib_irq_free(ctx, &adapter->irq);
2165 
2166 	for (int i = 0; i < adapter->rx_num_queues; i++, que++) {
2167 		iflib_irq_free(ctx, &que->que_irq);
2168 	}
2169 
2170 	/* First release all the interrupt resources */
2171 	if (adapter->memory != NULL) {
2172 		bus_release_resource(dev, SYS_RES_MEMORY,
2173 				     PCIR_BAR(0), adapter->memory);
2174 		adapter->memory = NULL;
2175 	}
2176 
2177 	if (adapter->flash != NULL) {
2178 		bus_release_resource(dev, SYS_RES_MEMORY,
2179 				     EM_FLASH, adapter->flash);
2180 		adapter->flash = NULL;
2181 	}
2182 	if (adapter->ioport != NULL)
2183 		bus_release_resource(dev, SYS_RES_IOPORT,
2184 		    adapter->io_rid, adapter->ioport);
2185 }
2186 
2187 /* Setup MSI or MSI/X */
2188 static int
2189 em_setup_msix(if_ctx_t ctx)
2190 {
2191 	struct adapter *adapter = iflib_get_softc(ctx);
2192 
2193 	if (adapter->hw.mac.type == e1000_82574) {
2194 		em_enable_vectors_82574(ctx);
2195 	}
2196 	return (0);
2197 }
2198 
2199 /*********************************************************************
2200  *
2201  *  Initialize the hardware to a configuration
2202  *  as specified by the adapter structure.
2203  *
2204  **********************************************************************/
2205 
2206 static void
2207 lem_smartspeed(struct adapter *adapter)
2208 {
2209 	u16 phy_tmp;
2210 
2211 	if (adapter->link_active || (adapter->hw.phy.type != e1000_phy_igp) ||
2212 	    adapter->hw.mac.autoneg == 0 ||
2213 	    (adapter->hw.phy.autoneg_advertised & ADVERTISE_1000_FULL) == 0)
2214 		return;
2215 
2216 	if (adapter->smartspeed == 0) {
2217 		/* If Master/Slave config fault is asserted twice,
2218 		 * we assume back-to-back */
2219 		e1000_read_phy_reg(&adapter->hw, PHY_1000T_STATUS, &phy_tmp);
2220 		if (!(phy_tmp & SR_1000T_MS_CONFIG_FAULT))
2221 			return;
2222 		e1000_read_phy_reg(&adapter->hw, PHY_1000T_STATUS, &phy_tmp);
2223 		if (phy_tmp & SR_1000T_MS_CONFIG_FAULT) {
2224 			e1000_read_phy_reg(&adapter->hw,
2225 			    PHY_1000T_CTRL, &phy_tmp);
2226 			if(phy_tmp & CR_1000T_MS_ENABLE) {
2227 				phy_tmp &= ~CR_1000T_MS_ENABLE;
2228 				e1000_write_phy_reg(&adapter->hw,
2229 				    PHY_1000T_CTRL, phy_tmp);
2230 				adapter->smartspeed++;
2231 				if(adapter->hw.mac.autoneg &&
2232 				   !e1000_copper_link_autoneg(&adapter->hw) &&
2233 				   !e1000_read_phy_reg(&adapter->hw,
2234 				    PHY_CONTROL, &phy_tmp)) {
2235 					phy_tmp |= (MII_CR_AUTO_NEG_EN |
2236 						    MII_CR_RESTART_AUTO_NEG);
2237 					e1000_write_phy_reg(&adapter->hw,
2238 					    PHY_CONTROL, phy_tmp);
2239 				}
2240 			}
2241 		}
2242 		return;
2243 	} else if(adapter->smartspeed == EM_SMARTSPEED_DOWNSHIFT) {
2244 		/* If still no link, perhaps using 2/3 pair cable */
2245 		e1000_read_phy_reg(&adapter->hw, PHY_1000T_CTRL, &phy_tmp);
2246 		phy_tmp |= CR_1000T_MS_ENABLE;
2247 		e1000_write_phy_reg(&adapter->hw, PHY_1000T_CTRL, phy_tmp);
2248 		if(adapter->hw.mac.autoneg &&
2249 		   !e1000_copper_link_autoneg(&adapter->hw) &&
2250 		   !e1000_read_phy_reg(&adapter->hw, PHY_CONTROL, &phy_tmp)) {
2251 			phy_tmp |= (MII_CR_AUTO_NEG_EN |
2252 				    MII_CR_RESTART_AUTO_NEG);
2253 			e1000_write_phy_reg(&adapter->hw, PHY_CONTROL, phy_tmp);
2254 		}
2255 	}
2256 	/* Restart process after EM_SMARTSPEED_MAX iterations */
2257 	if(adapter->smartspeed++ == EM_SMARTSPEED_MAX)
2258 		adapter->smartspeed = 0;
2259 }
2260 
2261 /*********************************************************************
2262  *
2263  *  Initialize the DMA Coalescing feature
2264  *
2265  **********************************************************************/
2266 static void
2267 igb_init_dmac(struct adapter *adapter, u32 pba)
2268 {
2269 	device_t	dev = adapter->dev;
2270 	struct e1000_hw *hw = &adapter->hw;
2271 	u32 		dmac, reg = ~E1000_DMACR_DMAC_EN;
2272 	u16		hwm;
2273 	u16		max_frame_size;
2274 
2275 	if (hw->mac.type == e1000_i211)
2276 		return;
2277 
2278 	max_frame_size = adapter->shared->isc_max_frame_size;
2279 	if (hw->mac.type > e1000_82580) {
2280 
2281 		if (adapter->dmac == 0) { /* Disabling it */
2282 			E1000_WRITE_REG(hw, E1000_DMACR, reg);
2283 			return;
2284 		} else
2285 			device_printf(dev, "DMA Coalescing enabled\n");
2286 
2287 		/* Set starting threshold */
2288 		E1000_WRITE_REG(hw, E1000_DMCTXTH, 0);
2289 
2290 		hwm = 64 * pba - max_frame_size / 16;
2291 		if (hwm < 64 * (pba - 6))
2292 			hwm = 64 * (pba - 6);
2293 		reg = E1000_READ_REG(hw, E1000_FCRTC);
2294 		reg &= ~E1000_FCRTC_RTH_COAL_MASK;
2295 		reg |= ((hwm << E1000_FCRTC_RTH_COAL_SHIFT)
2296 		    & E1000_FCRTC_RTH_COAL_MASK);
2297 		E1000_WRITE_REG(hw, E1000_FCRTC, reg);
2298 
2299 
2300 		dmac = pba - max_frame_size / 512;
2301 		if (dmac < pba - 10)
2302 			dmac = pba - 10;
2303 		reg = E1000_READ_REG(hw, E1000_DMACR);
2304 		reg &= ~E1000_DMACR_DMACTHR_MASK;
2305 		reg = ((dmac << E1000_DMACR_DMACTHR_SHIFT)
2306 		    & E1000_DMACR_DMACTHR_MASK);
2307 
2308 		/* transition to L0x or L1 if available..*/
2309 		reg |= (E1000_DMACR_DMAC_EN | E1000_DMACR_DMAC_LX_MASK);
2310 
2311 		/* Check if status is 2.5Gb backplane connection
2312 		* before configuration of watchdog timer, which is
2313 		* in msec values in 12.8usec intervals
2314 		* watchdog timer= msec values in 32usec intervals
2315 		* for non 2.5Gb connection
2316 		*/
2317 		if (hw->mac.type == e1000_i354) {
2318 			int status = E1000_READ_REG(hw, E1000_STATUS);
2319 			if ((status & E1000_STATUS_2P5_SKU) &&
2320 			    (!(status & E1000_STATUS_2P5_SKU_OVER)))
2321 				reg |= ((adapter->dmac * 5) >> 6);
2322 			else
2323 				reg |= (adapter->dmac >> 5);
2324 		} else {
2325 			reg |= (adapter->dmac >> 5);
2326 		}
2327 
2328 		E1000_WRITE_REG(hw, E1000_DMACR, reg);
2329 
2330 		E1000_WRITE_REG(hw, E1000_DMCRTRH, 0);
2331 
2332 		/* Set the interval before transition */
2333 		reg = E1000_READ_REG(hw, E1000_DMCTLX);
2334 		if (hw->mac.type == e1000_i350)
2335 			reg |= IGB_DMCTLX_DCFLUSH_DIS;
2336 		/*
2337 		** in 2.5Gb connection, TTLX unit is 0.4 usec
2338 		** which is 0x4*2 = 0xA. But delay is still 4 usec
2339 		*/
2340 		if (hw->mac.type == e1000_i354) {
2341 			int status = E1000_READ_REG(hw, E1000_STATUS);
2342 			if ((status & E1000_STATUS_2P5_SKU) &&
2343 			    (!(status & E1000_STATUS_2P5_SKU_OVER)))
2344 				reg |= 0xA;
2345 			else
2346 				reg |= 0x4;
2347 		} else {
2348 			reg |= 0x4;
2349 		}
2350 
2351 		E1000_WRITE_REG(hw, E1000_DMCTLX, reg);
2352 
2353 		/* free space in tx packet buffer to wake from DMA coal */
2354 		E1000_WRITE_REG(hw, E1000_DMCTXTH, (IGB_TXPBSIZE -
2355 		    (2 * max_frame_size)) >> 6);
2356 
2357 		/* make low power state decision controlled by DMA coal */
2358 		reg = E1000_READ_REG(hw, E1000_PCIEMISC);
2359 		reg &= ~E1000_PCIEMISC_LX_DECISION;
2360 		E1000_WRITE_REG(hw, E1000_PCIEMISC, reg);
2361 
2362 	} else if (hw->mac.type == e1000_82580) {
2363 		u32 reg = E1000_READ_REG(hw, E1000_PCIEMISC);
2364 		E1000_WRITE_REG(hw, E1000_PCIEMISC,
2365 		    reg & ~E1000_PCIEMISC_LX_DECISION);
2366 		E1000_WRITE_REG(hw, E1000_DMACR, 0);
2367 	}
2368 }
2369 
2370 static void
2371 em_reset(if_ctx_t ctx)
2372 {
2373 	device_t dev = iflib_get_dev(ctx);
2374 	struct adapter *adapter = iflib_get_softc(ctx);
2375 	struct ifnet *ifp = iflib_get_ifp(ctx);
2376 	struct e1000_hw *hw = &adapter->hw;
2377 	u16 rx_buffer_size;
2378 	u32 pba;
2379 
2380 	INIT_DEBUGOUT("em_reset: begin");
2381 	/* Let the firmware know the OS is in control */
2382 	em_get_hw_control(adapter);
2383 
2384 	/* Set up smart power down as default off on newer adapters. */
2385 	if (!em_smart_pwr_down && (hw->mac.type == e1000_82571 ||
2386 	    hw->mac.type == e1000_82572)) {
2387 		u16 phy_tmp = 0;
2388 
2389 		/* Speed up time to link by disabling smart power down. */
2390 		e1000_read_phy_reg(hw, IGP02E1000_PHY_POWER_MGMT, &phy_tmp);
2391 		phy_tmp &= ~IGP02E1000_PM_SPD;
2392 		e1000_write_phy_reg(hw, IGP02E1000_PHY_POWER_MGMT, phy_tmp);
2393 	}
2394 
2395 	/*
2396 	 * Packet Buffer Allocation (PBA)
2397 	 * Writing PBA sets the receive portion of the buffer
2398 	 * the remainder is used for the transmit buffer.
2399 	 */
2400 	switch (hw->mac.type) {
2401 	/* Total Packet Buffer on these is 48K */
2402 	case e1000_82571:
2403 	case e1000_82572:
2404 	case e1000_80003es2lan:
2405 			pba = E1000_PBA_32K; /* 32K for Rx, 16K for Tx */
2406 		break;
2407 	case e1000_82573: /* 82573: Total Packet Buffer is 32K */
2408 			pba = E1000_PBA_12K; /* 12K for Rx, 20K for Tx */
2409 		break;
2410 	case e1000_82574:
2411 	case e1000_82583:
2412 			pba = E1000_PBA_20K; /* 20K for Rx, 20K for Tx */
2413 		break;
2414 	case e1000_ich8lan:
2415 		pba = E1000_PBA_8K;
2416 		break;
2417 	case e1000_ich9lan:
2418 	case e1000_ich10lan:
2419 		/* Boost Receive side for jumbo frames */
2420 		if (adapter->hw.mac.max_frame_size > 4096)
2421 			pba = E1000_PBA_14K;
2422 		else
2423 			pba = E1000_PBA_10K;
2424 		break;
2425 	case e1000_pchlan:
2426 	case e1000_pch2lan:
2427 	case e1000_pch_lpt:
2428 	case e1000_pch_spt:
2429 	case e1000_pch_cnp:
2430 		pba = E1000_PBA_26K;
2431 		break;
2432 	case e1000_82575:
2433 		pba = E1000_PBA_32K;
2434 		break;
2435 	case e1000_82576:
2436 	case e1000_vfadapt:
2437 		pba = E1000_READ_REG(hw, E1000_RXPBS);
2438 		pba &= E1000_RXPBS_SIZE_MASK_82576;
2439 		break;
2440 	case e1000_82580:
2441 	case e1000_i350:
2442 	case e1000_i354:
2443 	case e1000_vfadapt_i350:
2444 		pba = E1000_READ_REG(hw, E1000_RXPBS);
2445 		pba = e1000_rxpbs_adjust_82580(pba);
2446 		break;
2447 	case e1000_i210:
2448 	case e1000_i211:
2449 		pba = E1000_PBA_34K;
2450 		break;
2451 	default:
2452 		if (adapter->hw.mac.max_frame_size > 8192)
2453 			pba = E1000_PBA_40K; /* 40K for Rx, 24K for Tx */
2454 		else
2455 			pba = E1000_PBA_48K; /* 48K for Rx, 16K for Tx */
2456 	}
2457 
2458 	/* Special needs in case of Jumbo frames */
2459 	if ((hw->mac.type == e1000_82575) && (ifp->if_mtu > ETHERMTU)) {
2460 		u32 tx_space, min_tx, min_rx;
2461 		pba = E1000_READ_REG(hw, E1000_PBA);
2462 		tx_space = pba >> 16;
2463 		pba &= 0xffff;
2464 		min_tx = (adapter->hw.mac.max_frame_size +
2465 		    sizeof(struct e1000_tx_desc) - ETHERNET_FCS_SIZE) * 2;
2466 		min_tx = roundup2(min_tx, 1024);
2467 		min_tx >>= 10;
2468 		min_rx = adapter->hw.mac.max_frame_size;
2469 		min_rx = roundup2(min_rx, 1024);
2470 		min_rx >>= 10;
2471 		if (tx_space < min_tx &&
2472 		    ((min_tx - tx_space) < pba)) {
2473 			pba = pba - (min_tx - tx_space);
2474 			/*
2475 			 * if short on rx space, rx wins
2476 			 * and must trump tx adjustment
2477 			 */
2478 			if (pba < min_rx)
2479 				pba = min_rx;
2480 		}
2481 		E1000_WRITE_REG(hw, E1000_PBA, pba);
2482 	}
2483 
2484 	if (hw->mac.type < igb_mac_min)
2485 		E1000_WRITE_REG(&adapter->hw, E1000_PBA, pba);
2486 
2487 	INIT_DEBUGOUT1("em_reset: pba=%dK",pba);
2488 
2489 	/*
2490 	 * These parameters control the automatic generation (Tx) and
2491 	 * response (Rx) to Ethernet PAUSE frames.
2492 	 * - High water mark should allow for at least two frames to be
2493 	 *   received after sending an XOFF.
2494 	 * - Low water mark works best when it is very near the high water mark.
2495 	 *   This allows the receiver to restart by sending XON when it has
2496 	 *   drained a bit. Here we use an arbitrary value of 1500 which will
2497 	 *   restart after one full frame is pulled from the buffer. There
2498 	 *   could be several smaller frames in the buffer and if so they will
2499 	 *   not trigger the XON until their total number reduces the buffer
2500 	 *   by 1500.
2501 	 * - The pause time is fairly large at 1000 x 512ns = 512 usec.
2502 	 */
2503 	rx_buffer_size = (pba & 0xffff) << 10;
2504 	hw->fc.high_water = rx_buffer_size -
2505 	    roundup2(adapter->hw.mac.max_frame_size, 1024);
2506 	hw->fc.low_water = hw->fc.high_water - 1500;
2507 
2508 	if (adapter->fc) /* locally set flow control value? */
2509 		hw->fc.requested_mode = adapter->fc;
2510 	else
2511 		hw->fc.requested_mode = e1000_fc_full;
2512 
2513 	if (hw->mac.type == e1000_80003es2lan)
2514 		hw->fc.pause_time = 0xFFFF;
2515 	else
2516 		hw->fc.pause_time = EM_FC_PAUSE_TIME;
2517 
2518 	hw->fc.send_xon = TRUE;
2519 
2520 	/* Device specific overrides/settings */
2521 	switch (hw->mac.type) {
2522 	case e1000_pchlan:
2523 		/* Workaround: no TX flow ctrl for PCH */
2524 		hw->fc.requested_mode = e1000_fc_rx_pause;
2525 		hw->fc.pause_time = 0xFFFF; /* override */
2526 		if (if_getmtu(ifp) > ETHERMTU) {
2527 			hw->fc.high_water = 0x3500;
2528 			hw->fc.low_water = 0x1500;
2529 		} else {
2530 			hw->fc.high_water = 0x5000;
2531 			hw->fc.low_water = 0x3000;
2532 		}
2533 		hw->fc.refresh_time = 0x1000;
2534 		break;
2535 	case e1000_pch2lan:
2536 	case e1000_pch_lpt:
2537 	case e1000_pch_spt:
2538 	case e1000_pch_cnp:
2539 		hw->fc.high_water = 0x5C20;
2540 		hw->fc.low_water = 0x5048;
2541 		hw->fc.pause_time = 0x0650;
2542 		hw->fc.refresh_time = 0x0400;
2543 		/* Jumbos need adjusted PBA */
2544 		if (if_getmtu(ifp) > ETHERMTU)
2545 			E1000_WRITE_REG(hw, E1000_PBA, 12);
2546 		else
2547 			E1000_WRITE_REG(hw, E1000_PBA, 26);
2548 		break;
2549 	case e1000_82575:
2550 	case e1000_82576:
2551 		/* 8-byte granularity */
2552 		hw->fc.low_water = hw->fc.high_water - 8;
2553 		break;
2554 	case e1000_82580:
2555 	case e1000_i350:
2556 	case e1000_i354:
2557 	case e1000_i210:
2558 	case e1000_i211:
2559 	case e1000_vfadapt:
2560 	case e1000_vfadapt_i350:
2561 		/* 16-byte granularity */
2562 		hw->fc.low_water = hw->fc.high_water - 16;
2563 		break;
2564 	case e1000_ich9lan:
2565 	case e1000_ich10lan:
2566 		if (if_getmtu(ifp) > ETHERMTU) {
2567 			hw->fc.high_water = 0x2800;
2568 			hw->fc.low_water = hw->fc.high_water - 8;
2569 			break;
2570 		}
2571 		/* FALLTHROUGH */
2572 	default:
2573 		if (hw->mac.type == e1000_80003es2lan)
2574 			hw->fc.pause_time = 0xFFFF;
2575 		break;
2576 	}
2577 
2578 	/* Issue a global reset */
2579 	e1000_reset_hw(hw);
2580 	if (adapter->hw.mac.type >= igb_mac_min) {
2581 		E1000_WRITE_REG(hw, E1000_WUC, 0);
2582 	} else {
2583 		E1000_WRITE_REG(hw, E1000_WUFC, 0);
2584 		em_disable_aspm(adapter);
2585 	}
2586 	if (adapter->flags & IGB_MEDIA_RESET) {
2587 		e1000_setup_init_funcs(hw, TRUE);
2588 		e1000_get_bus_info(hw);
2589 		adapter->flags &= ~IGB_MEDIA_RESET;
2590 	}
2591 	/* and a re-init */
2592 	if (e1000_init_hw(hw) < 0) {
2593 		device_printf(dev, "Hardware Initialization Failed\n");
2594 		return;
2595 	}
2596 	if (adapter->hw.mac.type >= igb_mac_min)
2597 		igb_init_dmac(adapter, pba);
2598 
2599 	E1000_WRITE_REG(hw, E1000_VET, ETHERTYPE_VLAN);
2600 	e1000_get_phy_info(hw);
2601 	e1000_check_for_link(hw);
2602 }
2603 
2604 #define RSSKEYLEN 10
2605 static void
2606 em_initialize_rss_mapping(struct adapter *adapter)
2607 {
2608 	uint8_t  rss_key[4 * RSSKEYLEN];
2609 	uint32_t reta = 0;
2610 	struct e1000_hw	*hw = &adapter->hw;
2611 	int i;
2612 
2613 	/*
2614 	 * Configure RSS key
2615 	 */
2616 	arc4rand(rss_key, sizeof(rss_key), 0);
2617 	for (i = 0; i < RSSKEYLEN; ++i) {
2618 		uint32_t rssrk = 0;
2619 
2620 		rssrk = EM_RSSRK_VAL(rss_key, i);
2621 		E1000_WRITE_REG(hw,E1000_RSSRK(i), rssrk);
2622 	}
2623 
2624 	/*
2625 	 * Configure RSS redirect table in following fashion:
2626 	 * (hash & ring_cnt_mask) == rdr_table[(hash & rdr_table_mask)]
2627 	 */
2628 	for (i = 0; i < sizeof(reta); ++i) {
2629 		uint32_t q;
2630 
2631 		q = (i % adapter->rx_num_queues) << 7;
2632 		reta |= q << (8 * i);
2633 	}
2634 
2635 	for (i = 0; i < 32; ++i)
2636 		E1000_WRITE_REG(hw, E1000_RETA(i), reta);
2637 
2638 	E1000_WRITE_REG(hw, E1000_MRQC, E1000_MRQC_RSS_ENABLE_2Q |
2639 			E1000_MRQC_RSS_FIELD_IPV4_TCP |
2640 			E1000_MRQC_RSS_FIELD_IPV4 |
2641 			E1000_MRQC_RSS_FIELD_IPV6_TCP_EX |
2642 			E1000_MRQC_RSS_FIELD_IPV6_EX |
2643 			E1000_MRQC_RSS_FIELD_IPV6);
2644 
2645 }
2646 
2647 static void
2648 igb_initialize_rss_mapping(struct adapter *adapter)
2649 {
2650 	struct e1000_hw *hw = &adapter->hw;
2651 	int i;
2652 	int queue_id;
2653 	u32 reta;
2654 	u32 rss_key[10], mrqc, shift = 0;
2655 
2656 	/* XXX? */
2657 	if (adapter->hw.mac.type == e1000_82575)
2658 		shift = 6;
2659 
2660 	/*
2661 	 * The redirection table controls which destination
2662 	 * queue each bucket redirects traffic to.
2663 	 * Each DWORD represents four queues, with the LSB
2664 	 * being the first queue in the DWORD.
2665 	 *
2666 	 * This just allocates buckets to queues using round-robin
2667 	 * allocation.
2668 	 *
2669 	 * NOTE: It Just Happens to line up with the default
2670 	 * RSS allocation method.
2671 	 */
2672 
2673 	/* Warning FM follows */
2674 	reta = 0;
2675 	for (i = 0; i < 128; i++) {
2676 #ifdef RSS
2677 		queue_id = rss_get_indirection_to_bucket(i);
2678 		/*
2679 		 * If we have more queues than buckets, we'll
2680 		 * end up mapping buckets to a subset of the
2681 		 * queues.
2682 		 *
2683 		 * If we have more buckets than queues, we'll
2684 		 * end up instead assigning multiple buckets
2685 		 * to queues.
2686 		 *
2687 		 * Both are suboptimal, but we need to handle
2688 		 * the case so we don't go out of bounds
2689 		 * indexing arrays and such.
2690 		 */
2691 		queue_id = queue_id % adapter->rx_num_queues;
2692 #else
2693 		queue_id = (i % adapter->rx_num_queues);
2694 #endif
2695 		/* Adjust if required */
2696 		queue_id = queue_id << shift;
2697 
2698 		/*
2699 		 * The low 8 bits are for hash value (n+0);
2700 		 * The next 8 bits are for hash value (n+1), etc.
2701 		 */
2702 		reta = reta >> 8;
2703 		reta = reta | ( ((uint32_t) queue_id) << 24);
2704 		if ((i & 3) == 3) {
2705 			E1000_WRITE_REG(hw, E1000_RETA(i >> 2), reta);
2706 			reta = 0;
2707 		}
2708 	}
2709 
2710 	/* Now fill in hash table */
2711 
2712 	/*
2713 	 * MRQC: Multiple Receive Queues Command
2714 	 * Set queuing to RSS control, number depends on the device.
2715 	 */
2716 	mrqc = E1000_MRQC_ENABLE_RSS_8Q;
2717 
2718 #ifdef RSS
2719 	/* XXX ew typecasting */
2720 	rss_getkey((uint8_t *) &rss_key);
2721 #else
2722 	arc4rand(&rss_key, sizeof(rss_key), 0);
2723 #endif
2724 	for (i = 0; i < 10; i++)
2725 		E1000_WRITE_REG_ARRAY(hw, E1000_RSSRK(0), i, rss_key[i]);
2726 
2727 	/*
2728 	 * Configure the RSS fields to hash upon.
2729 	 */
2730 	mrqc |= (E1000_MRQC_RSS_FIELD_IPV4 |
2731 	    E1000_MRQC_RSS_FIELD_IPV4_TCP);
2732 	mrqc |= (E1000_MRQC_RSS_FIELD_IPV6 |
2733 	    E1000_MRQC_RSS_FIELD_IPV6_TCP);
2734 	mrqc |=( E1000_MRQC_RSS_FIELD_IPV4_UDP |
2735 	    E1000_MRQC_RSS_FIELD_IPV6_UDP);
2736 	mrqc |=( E1000_MRQC_RSS_FIELD_IPV6_UDP_EX |
2737 	    E1000_MRQC_RSS_FIELD_IPV6_TCP_EX);
2738 
2739 	E1000_WRITE_REG(hw, E1000_MRQC, mrqc);
2740 }
2741 
2742 /*********************************************************************
2743  *
2744  *  Setup networking device structure and register an interface.
2745  *
2746  **********************************************************************/
2747 static int
2748 em_setup_interface(if_ctx_t ctx)
2749 {
2750 	struct ifnet *ifp = iflib_get_ifp(ctx);
2751 	struct adapter *adapter = iflib_get_softc(ctx);
2752 	if_softc_ctx_t scctx = adapter->shared;
2753 	uint64_t cap = 0;
2754 
2755 	INIT_DEBUGOUT("em_setup_interface: begin");
2756 
2757 	/* TSO parameters */
2758 	if_sethwtsomax(ifp, IP_MAXPACKET);
2759 	/* Take m_pullup(9)'s in em_xmit() w/ TSO into acount. */
2760 	if_sethwtsomaxsegcount(ifp, EM_MAX_SCATTER - 5);
2761 	if_sethwtsomaxsegsize(ifp, EM_TSO_SEG_SIZE);
2762 
2763 	/* Single Queue */
2764 	if (adapter->tx_num_queues == 1) {
2765 		if_setsendqlen(ifp, scctx->isc_ntxd[0] - 1);
2766 		if_setsendqready(ifp);
2767 	}
2768 
2769 	cap = IFCAP_HWCSUM | IFCAP_VLAN_HWCSUM | IFCAP_TSO4;
2770 	cap |= IFCAP_VLAN_HWTAGGING | IFCAP_VLAN_HWTSO | IFCAP_VLAN_MTU;
2771 
2772 	/*
2773 	 * Tell the upper layer(s) we
2774 	 * support full VLAN capability
2775 	 */
2776 	if_setifheaderlen(ifp, sizeof(struct ether_vlan_header));
2777 	if_setcapabilitiesbit(ifp, cap, 0);
2778 
2779 	/*
2780 	 * Don't turn this on by default, if vlans are
2781 	 * created on another pseudo device (eg. lagg)
2782 	 * then vlan events are not passed thru, breaking
2783 	 * operation, but with HW FILTER off it works. If
2784 	 * using vlans directly on the em driver you can
2785 	 * enable this and get full hardware tag filtering.
2786 	 */
2787 	if_setcapabilitiesbit(ifp, IFCAP_VLAN_HWFILTER,0);
2788 
2789 	/* Enable only WOL MAGIC by default */
2790 	if (adapter->wol) {
2791 		if_setcapenablebit(ifp, IFCAP_WOL_MAGIC,
2792 			    IFCAP_WOL_MCAST| IFCAP_WOL_UCAST);
2793 	} else {
2794 		if_setcapenablebit(ifp, 0, IFCAP_WOL_MAGIC |
2795 			     IFCAP_WOL_MCAST| IFCAP_WOL_UCAST);
2796 	}
2797 
2798 	/*
2799 	 * Specify the media types supported by this adapter and register
2800 	 * callbacks to update media and link information
2801 	 */
2802 	if ((adapter->hw.phy.media_type == e1000_media_type_fiber) ||
2803 	    (adapter->hw.phy.media_type == e1000_media_type_internal_serdes)) {
2804 		u_char fiber_type = IFM_1000_SX;	/* default type */
2805 
2806 		if (adapter->hw.mac.type == e1000_82545)
2807 			fiber_type = IFM_1000_LX;
2808 		ifmedia_add(adapter->media, IFM_ETHER | fiber_type | IFM_FDX, 0, NULL);
2809 		ifmedia_add(adapter->media, IFM_ETHER | fiber_type, 0, NULL);
2810 	} else {
2811 		ifmedia_add(adapter->media, IFM_ETHER | IFM_10_T, 0, NULL);
2812 		ifmedia_add(adapter->media, IFM_ETHER | IFM_10_T | IFM_FDX, 0, NULL);
2813 		ifmedia_add(adapter->media, IFM_ETHER | IFM_100_TX, 0, NULL);
2814 		ifmedia_add(adapter->media, IFM_ETHER | IFM_100_TX | IFM_FDX, 0, NULL);
2815 		if (adapter->hw.phy.type != e1000_phy_ife) {
2816 			ifmedia_add(adapter->media, IFM_ETHER | IFM_1000_T | IFM_FDX, 0, NULL);
2817 			ifmedia_add(adapter->media, IFM_ETHER | IFM_1000_T, 0, NULL);
2818 		}
2819 	}
2820 	ifmedia_add(adapter->media, IFM_ETHER | IFM_AUTO, 0, NULL);
2821 	ifmedia_set(adapter->media, IFM_ETHER | IFM_AUTO);
2822 	return (0);
2823 }
2824 
2825 static int
2826 em_if_tx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int ntxqs, int ntxqsets)
2827 {
2828 	struct adapter *adapter = iflib_get_softc(ctx);
2829 	if_softc_ctx_t scctx = adapter->shared;
2830 	int error = E1000_SUCCESS;
2831 	struct em_tx_queue *que;
2832 	int i, j;
2833 
2834 	MPASS(adapter->tx_num_queues > 0);
2835 	MPASS(adapter->tx_num_queues == ntxqsets);
2836 
2837 	/* First allocate the top level queue structs */
2838 	if (!(adapter->tx_queues =
2839 	    (struct em_tx_queue *) malloc(sizeof(struct em_tx_queue) *
2840 	    adapter->tx_num_queues, M_DEVBUF, M_NOWAIT | M_ZERO))) {
2841 		device_printf(iflib_get_dev(ctx), "Unable to allocate queue memory\n");
2842 		return(ENOMEM);
2843 	}
2844 
2845 	for (i = 0, que = adapter->tx_queues; i < adapter->tx_num_queues; i++, que++) {
2846 		/* Set up some basics */
2847 
2848 		struct tx_ring *txr = &que->txr;
2849 		txr->adapter = que->adapter = adapter;
2850 		que->me = txr->me =  i;
2851 
2852 		/* Allocate report status array */
2853 		if (!(txr->tx_rsq = (qidx_t *) malloc(sizeof(qidx_t) * scctx->isc_ntxd[0], M_DEVBUF, M_NOWAIT | M_ZERO))) {
2854 			device_printf(iflib_get_dev(ctx), "failed to allocate rs_idxs memory\n");
2855 			error = ENOMEM;
2856 			goto fail;
2857 		}
2858 		for (j = 0; j < scctx->isc_ntxd[0]; j++)
2859 			txr->tx_rsq[j] = QIDX_INVALID;
2860 		/* get the virtual and physical address of the hardware queues */
2861 		txr->tx_base = (struct e1000_tx_desc *)vaddrs[i*ntxqs];
2862 		txr->tx_paddr = paddrs[i*ntxqs];
2863 	}
2864 
2865 	device_printf(iflib_get_dev(ctx), "allocated for %d tx_queues\n", adapter->tx_num_queues);
2866 	return (0);
2867 fail:
2868 	em_if_queues_free(ctx);
2869 	return (error);
2870 }
2871 
2872 static int
2873 em_if_rx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int nrxqs, int nrxqsets)
2874 {
2875 	struct adapter *adapter = iflib_get_softc(ctx);
2876 	int error = E1000_SUCCESS;
2877 	struct em_rx_queue *que;
2878 	int i;
2879 
2880 	MPASS(adapter->rx_num_queues > 0);
2881 	MPASS(adapter->rx_num_queues == nrxqsets);
2882 
2883 	/* First allocate the top level queue structs */
2884 	if (!(adapter->rx_queues =
2885 	    (struct em_rx_queue *) malloc(sizeof(struct em_rx_queue) *
2886 	    adapter->rx_num_queues, M_DEVBUF, M_NOWAIT | M_ZERO))) {
2887 		device_printf(iflib_get_dev(ctx), "Unable to allocate queue memory\n");
2888 		error = ENOMEM;
2889 		goto fail;
2890 	}
2891 
2892 	for (i = 0, que = adapter->rx_queues; i < nrxqsets; i++, que++) {
2893 		/* Set up some basics */
2894 		struct rx_ring *rxr = &que->rxr;
2895 		rxr->adapter = que->adapter = adapter;
2896 		rxr->que = que;
2897 		que->me = rxr->me =  i;
2898 
2899 		/* get the virtual and physical address of the hardware queues */
2900 		rxr->rx_base = (union e1000_rx_desc_extended *)vaddrs[i*nrxqs];
2901 		rxr->rx_paddr = paddrs[i*nrxqs];
2902 	}
2903 
2904 	device_printf(iflib_get_dev(ctx), "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 MSIX 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 MSIX IRQ Handled");
4022 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "mbuf_defrag_fail",
4023 			 CTLFLAG_RD, &adapter->mbuf_defrag_failed,
4024 			 "Defragmenting mbuf chain failed");
4025 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "tx_dma_fail",
4026 			CTLFLAG_RD, &adapter->no_tx_dma_setup,
4027 			"Driver tx dma failure in xmit");
4028 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "rx_overruns",
4029 			CTLFLAG_RD, &adapter->rx_overruns,
4030 			"RX overruns");
4031 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "watchdog_timeouts",
4032 			CTLFLAG_RD, &adapter->watchdog_events,
4033 			"Watchdog timeouts");
4034 	SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "device_control",
4035 			CTLTYPE_UINT | CTLFLAG_RD, adapter, E1000_CTRL,
4036 			em_sysctl_reg_handler, "IU",
4037 			"Device Control Register");
4038 	SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "rx_control",
4039 			CTLTYPE_UINT | CTLFLAG_RD, adapter, E1000_RCTL,
4040 			em_sysctl_reg_handler, "IU",
4041 			"Receiver Control Register");
4042 	SYSCTL_ADD_UINT(ctx, child, OID_AUTO, "fc_high_water",
4043 			CTLFLAG_RD, &adapter->hw.fc.high_water, 0,
4044 			"Flow Control High Watermark");
4045 	SYSCTL_ADD_UINT(ctx, child, OID_AUTO, "fc_low_water",
4046 			CTLFLAG_RD, &adapter->hw.fc.low_water, 0,
4047 			"Flow Control Low Watermark");
4048 
4049 	for (int i = 0; i < adapter->tx_num_queues; i++, tx_que++) {
4050 		struct tx_ring *txr = &tx_que->txr;
4051 		snprintf(namebuf, QUEUE_NAME_LEN, "queue_tx_%d", i);
4052 		queue_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, namebuf,
4053 					    CTLFLAG_RD, NULL, "TX Queue Name");
4054 		queue_list = SYSCTL_CHILDREN(queue_node);
4055 
4056 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "txd_head",
4057 				CTLTYPE_UINT | CTLFLAG_RD, adapter,
4058 				E1000_TDH(txr->me),
4059 				em_sysctl_reg_handler, "IU",
4060 				"Transmit Descriptor Head");
4061 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "txd_tail",
4062 				CTLTYPE_UINT | CTLFLAG_RD, adapter,
4063 				E1000_TDT(txr->me),
4064 				em_sysctl_reg_handler, "IU",
4065 				"Transmit Descriptor Tail");
4066 		SYSCTL_ADD_ULONG(ctx, queue_list, OID_AUTO, "tx_irq",
4067 				CTLFLAG_RD, &txr->tx_irq,
4068 				"Queue MSI-X Transmit Interrupts");
4069 	}
4070 
4071 	for (int j = 0; j < adapter->rx_num_queues; j++, rx_que++) {
4072 		struct rx_ring *rxr = &rx_que->rxr;
4073 		snprintf(namebuf, QUEUE_NAME_LEN, "queue_rx_%d", j);
4074 		queue_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, namebuf,
4075 					    CTLFLAG_RD, NULL, "RX Queue Name");
4076 		queue_list = SYSCTL_CHILDREN(queue_node);
4077 
4078 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "rxd_head",
4079 				CTLTYPE_UINT | CTLFLAG_RD, adapter,
4080 				E1000_RDH(rxr->me),
4081 				em_sysctl_reg_handler, "IU",
4082 				"Receive Descriptor Head");
4083 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "rxd_tail",
4084 				CTLTYPE_UINT | CTLFLAG_RD, adapter,
4085 				E1000_RDT(rxr->me),
4086 				em_sysctl_reg_handler, "IU",
4087 				"Receive Descriptor Tail");
4088 		SYSCTL_ADD_ULONG(ctx, queue_list, OID_AUTO, "rx_irq",
4089 				CTLFLAG_RD, &rxr->rx_irq,
4090 				"Queue MSI-X Receive Interrupts");
4091 	}
4092 
4093 	/* MAC stats get their own sub node */
4094 
4095 	stat_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, "mac_stats",
4096 				    CTLFLAG_RD, NULL, "Statistics");
4097 	stat_list = SYSCTL_CHILDREN(stat_node);
4098 
4099 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "excess_coll",
4100 			CTLFLAG_RD, &stats->ecol,
4101 			"Excessive collisions");
4102 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "single_coll",
4103 			CTLFLAG_RD, &stats->scc,
4104 			"Single collisions");
4105 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "multiple_coll",
4106 			CTLFLAG_RD, &stats->mcc,
4107 			"Multiple collisions");
4108 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "late_coll",
4109 			CTLFLAG_RD, &stats->latecol,
4110 			"Late collisions");
4111 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "collision_count",
4112 			CTLFLAG_RD, &stats->colc,
4113 			"Collision Count");
4114 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "symbol_errors",
4115 			CTLFLAG_RD, &adapter->stats.symerrs,
4116 			"Symbol Errors");
4117 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "sequence_errors",
4118 			CTLFLAG_RD, &adapter->stats.sec,
4119 			"Sequence Errors");
4120 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "defer_count",
4121 			CTLFLAG_RD, &adapter->stats.dc,
4122 			"Defer Count");
4123 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "missed_packets",
4124 			CTLFLAG_RD, &adapter->stats.mpc,
4125 			"Missed Packets");
4126 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_no_buff",
4127 			CTLFLAG_RD, &adapter->stats.rnbc,
4128 			"Receive No Buffers");
4129 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_undersize",
4130 			CTLFLAG_RD, &adapter->stats.ruc,
4131 			"Receive Undersize");
4132 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_fragmented",
4133 			CTLFLAG_RD, &adapter->stats.rfc,
4134 			"Fragmented Packets Received ");
4135 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_oversize",
4136 			CTLFLAG_RD, &adapter->stats.roc,
4137 			"Oversized Packets Received");
4138 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_jabber",
4139 			CTLFLAG_RD, &adapter->stats.rjc,
4140 			"Recevied Jabber");
4141 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_errs",
4142 			CTLFLAG_RD, &adapter->stats.rxerrc,
4143 			"Receive Errors");
4144 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "crc_errs",
4145 			CTLFLAG_RD, &adapter->stats.crcerrs,
4146 			"CRC errors");
4147 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "alignment_errs",
4148 			CTLFLAG_RD, &adapter->stats.algnerrc,
4149 			"Alignment Errors");
4150 	/* On 82575 these are collision counts */
4151 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "coll_ext_errs",
4152 			CTLFLAG_RD, &adapter->stats.cexterr,
4153 			"Collision/Carrier extension errors");
4154 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xon_recvd",
4155 			CTLFLAG_RD, &adapter->stats.xonrxc,
4156 			"XON Received");
4157 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xon_txd",
4158 			CTLFLAG_RD, &adapter->stats.xontxc,
4159 			"XON Transmitted");
4160 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xoff_recvd",
4161 			CTLFLAG_RD, &adapter->stats.xoffrxc,
4162 			"XOFF Received");
4163 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xoff_txd",
4164 			CTLFLAG_RD, &adapter->stats.xofftxc,
4165 			"XOFF Transmitted");
4166 
4167 	/* Packet Reception Stats */
4168 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_pkts_recvd",
4169 			CTLFLAG_RD, &adapter->stats.tpr,
4170 			"Total Packets Received ");
4171 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_pkts_recvd",
4172 			CTLFLAG_RD, &adapter->stats.gprc,
4173 			"Good Packets Received");
4174 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "bcast_pkts_recvd",
4175 			CTLFLAG_RD, &adapter->stats.bprc,
4176 			"Broadcast Packets Received");
4177 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "mcast_pkts_recvd",
4178 			CTLFLAG_RD, &adapter->stats.mprc,
4179 			"Multicast Packets Received");
4180 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_64",
4181 			CTLFLAG_RD, &adapter->stats.prc64,
4182 			"64 byte frames received ");
4183 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_65_127",
4184 			CTLFLAG_RD, &adapter->stats.prc127,
4185 			"65-127 byte frames received");
4186 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_128_255",
4187 			CTLFLAG_RD, &adapter->stats.prc255,
4188 			"128-255 byte frames received");
4189 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_256_511",
4190 			CTLFLAG_RD, &adapter->stats.prc511,
4191 			"256-511 byte frames received");
4192 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_512_1023",
4193 			CTLFLAG_RD, &adapter->stats.prc1023,
4194 			"512-1023 byte frames received");
4195 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_1024_1522",
4196 			CTLFLAG_RD, &adapter->stats.prc1522,
4197 			"1023-1522 byte frames received");
4198 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_octets_recvd",
4199 			CTLFLAG_RD, &adapter->stats.gorc,
4200 			"Good Octets Received");
4201 
4202 	/* Packet Transmission Stats */
4203 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_octets_txd",
4204 			CTLFLAG_RD, &adapter->stats.gotc,
4205 			"Good Octets Transmitted");
4206 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_pkts_txd",
4207 			CTLFLAG_RD, &adapter->stats.tpt,
4208 			"Total Packets Transmitted");
4209 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_pkts_txd",
4210 			CTLFLAG_RD, &adapter->stats.gptc,
4211 			"Good Packets Transmitted");
4212 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "bcast_pkts_txd",
4213 			CTLFLAG_RD, &adapter->stats.bptc,
4214 			"Broadcast Packets Transmitted");
4215 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "mcast_pkts_txd",
4216 			CTLFLAG_RD, &adapter->stats.mptc,
4217 			"Multicast Packets Transmitted");
4218 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_64",
4219 			CTLFLAG_RD, &adapter->stats.ptc64,
4220 			"64 byte frames transmitted ");
4221 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_65_127",
4222 			CTLFLAG_RD, &adapter->stats.ptc127,
4223 			"65-127 byte frames transmitted");
4224 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_128_255",
4225 			CTLFLAG_RD, &adapter->stats.ptc255,
4226 			"128-255 byte frames transmitted");
4227 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_256_511",
4228 			CTLFLAG_RD, &adapter->stats.ptc511,
4229 			"256-511 byte frames transmitted");
4230 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_512_1023",
4231 			CTLFLAG_RD, &adapter->stats.ptc1023,
4232 			"512-1023 byte frames transmitted");
4233 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_1024_1522",
4234 			CTLFLAG_RD, &adapter->stats.ptc1522,
4235 			"1024-1522 byte frames transmitted");
4236 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tso_txd",
4237 			CTLFLAG_RD, &adapter->stats.tsctc,
4238 			"TSO Contexts Transmitted");
4239 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tso_ctx_fail",
4240 			CTLFLAG_RD, &adapter->stats.tsctfc,
4241 			"TSO Contexts Failed");
4242 
4243 
4244 	/* Interrupt Stats */
4245 
4246 	int_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, "interrupts",
4247 				    CTLFLAG_RD, NULL, "Interrupt Statistics");
4248 	int_list = SYSCTL_CHILDREN(int_node);
4249 
4250 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "asserts",
4251 			CTLFLAG_RD, &adapter->stats.iac,
4252 			"Interrupt Assertion Count");
4253 
4254 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_pkt_timer",
4255 			CTLFLAG_RD, &adapter->stats.icrxptc,
4256 			"Interrupt Cause Rx Pkt Timer Expire Count");
4257 
4258 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_abs_timer",
4259 			CTLFLAG_RD, &adapter->stats.icrxatc,
4260 			"Interrupt Cause Rx Abs Timer Expire Count");
4261 
4262 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_pkt_timer",
4263 			CTLFLAG_RD, &adapter->stats.ictxptc,
4264 			"Interrupt Cause Tx Pkt Timer Expire Count");
4265 
4266 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_abs_timer",
4267 			CTLFLAG_RD, &adapter->stats.ictxatc,
4268 			"Interrupt Cause Tx Abs Timer Expire Count");
4269 
4270 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_queue_empty",
4271 			CTLFLAG_RD, &adapter->stats.ictxqec,
4272 			"Interrupt Cause Tx Queue Empty Count");
4273 
4274 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_queue_min_thresh",
4275 			CTLFLAG_RD, &adapter->stats.ictxqmtc,
4276 			"Interrupt Cause Tx Queue Min Thresh Count");
4277 
4278 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_desc_min_thresh",
4279 			CTLFLAG_RD, &adapter->stats.icrxdmtc,
4280 			"Interrupt Cause Rx Desc Min Thresh Count");
4281 
4282 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_overrun",
4283 			CTLFLAG_RD, &adapter->stats.icrxoc,
4284 			"Interrupt Cause Receiver Overrun Count");
4285 }
4286 
4287 /**********************************************************************
4288  *
4289  *  This routine provides a way to dump out the adapter eeprom,
4290  *  often a useful debug/service tool. This only dumps the first
4291  *  32 words, stuff that matters is in that extent.
4292  *
4293  **********************************************************************/
4294 static int
4295 em_sysctl_nvm_info(SYSCTL_HANDLER_ARGS)
4296 {
4297 	struct adapter *adapter = (struct adapter *)arg1;
4298 	int error;
4299 	int result;
4300 
4301 	result = -1;
4302 	error = sysctl_handle_int(oidp, &result, 0, req);
4303 
4304 	if (error || !req->newptr)
4305 		return (error);
4306 
4307 	/*
4308 	 * This value will cause a hex dump of the
4309 	 * first 32 16-bit words of the EEPROM to
4310 	 * the screen.
4311 	 */
4312 	if (result == 1)
4313 		em_print_nvm_info(adapter);
4314 
4315 	return (error);
4316 }
4317 
4318 static void
4319 em_print_nvm_info(struct adapter *adapter)
4320 {
4321 	u16 eeprom_data;
4322 	int i, j, row = 0;
4323 
4324 	/* Its a bit crude, but it gets the job done */
4325 	printf("\nInterface EEPROM Dump:\n");
4326 	printf("Offset\n0x0000  ");
4327 	for (i = 0, j = 0; i < 32; i++, j++) {
4328 		if (j == 8) { /* Make the offset block */
4329 			j = 0; ++row;
4330 			printf("\n0x00%x0  ",row);
4331 		}
4332 		e1000_read_nvm(&adapter->hw, i, 1, &eeprom_data);
4333 		printf("%04x ", eeprom_data);
4334 	}
4335 	printf("\n");
4336 }
4337 
4338 static int
4339 em_sysctl_int_delay(SYSCTL_HANDLER_ARGS)
4340 {
4341 	struct em_int_delay_info *info;
4342 	struct adapter *adapter;
4343 	u32 regval;
4344 	int error, usecs, ticks;
4345 
4346 	info = (struct em_int_delay_info *) arg1;
4347 	usecs = info->value;
4348 	error = sysctl_handle_int(oidp, &usecs, 0, req);
4349 	if (error != 0 || req->newptr == NULL)
4350 		return (error);
4351 	if (usecs < 0 || usecs > EM_TICKS_TO_USECS(65535))
4352 		return (EINVAL);
4353 	info->value = usecs;
4354 	ticks = EM_USECS_TO_TICKS(usecs);
4355 	if (info->offset == E1000_ITR)	/* units are 256ns here */
4356 		ticks *= 4;
4357 
4358 	adapter = info->adapter;
4359 
4360 	regval = E1000_READ_OFFSET(&adapter->hw, info->offset);
4361 	regval = (regval & ~0xffff) | (ticks & 0xffff);
4362 	/* Handle a few special cases. */
4363 	switch (info->offset) {
4364 	case E1000_RDTR:
4365 		break;
4366 	case E1000_TIDV:
4367 		if (ticks == 0) {
4368 			adapter->txd_cmd &= ~E1000_TXD_CMD_IDE;
4369 			/* Don't write 0 into the TIDV register. */
4370 			regval++;
4371 		} else
4372 			adapter->txd_cmd |= E1000_TXD_CMD_IDE;
4373 		break;
4374 	}
4375 	E1000_WRITE_OFFSET(&adapter->hw, info->offset, regval);
4376 	return (0);
4377 }
4378 
4379 static void
4380 em_add_int_delay_sysctl(struct adapter *adapter, const char *name,
4381 	const char *description, struct em_int_delay_info *info,
4382 	int offset, int value)
4383 {
4384 	info->adapter = adapter;
4385 	info->offset = offset;
4386 	info->value = value;
4387 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(adapter->dev),
4388 	    SYSCTL_CHILDREN(device_get_sysctl_tree(adapter->dev)),
4389 	    OID_AUTO, name, CTLTYPE_INT|CTLFLAG_RW,
4390 	    info, 0, em_sysctl_int_delay, "I", description);
4391 }
4392 
4393 /*
4394  * Set flow control using sysctl:
4395  * Flow control values:
4396  *      0 - off
4397  *      1 - rx pause
4398  *      2 - tx pause
4399  *      3 - full
4400  */
4401 static int
4402 em_set_flowcntl(SYSCTL_HANDLER_ARGS)
4403 {
4404 	int error;
4405 	static int input = 3; /* default is full */
4406 	struct adapter	*adapter = (struct adapter *) arg1;
4407 
4408 	error = sysctl_handle_int(oidp, &input, 0, req);
4409 
4410 	if ((error) || (req->newptr == NULL))
4411 		return (error);
4412 
4413 	if (input == adapter->fc) /* no change? */
4414 		return (error);
4415 
4416 	switch (input) {
4417 	case e1000_fc_rx_pause:
4418 	case e1000_fc_tx_pause:
4419 	case e1000_fc_full:
4420 	case e1000_fc_none:
4421 		adapter->hw.fc.requested_mode = input;
4422 		adapter->fc = input;
4423 		break;
4424 	default:
4425 		/* Do nothing */
4426 		return (error);
4427 	}
4428 
4429 	adapter->hw.fc.current_mode = adapter->hw.fc.requested_mode;
4430 	e1000_force_mac_fc(&adapter->hw);
4431 	return (error);
4432 }
4433 
4434 /*
4435  * Manage Energy Efficient Ethernet:
4436  * Control values:
4437  *     0/1 - enabled/disabled
4438  */
4439 static int
4440 em_sysctl_eee(SYSCTL_HANDLER_ARGS)
4441 {
4442 	struct adapter *adapter = (struct adapter *) arg1;
4443 	int error, value;
4444 
4445 	value = adapter->hw.dev_spec.ich8lan.eee_disable;
4446 	error = sysctl_handle_int(oidp, &value, 0, req);
4447 	if (error || req->newptr == NULL)
4448 		return (error);
4449 	adapter->hw.dev_spec.ich8lan.eee_disable = (value != 0);
4450 	em_if_init(adapter->ctx);
4451 
4452 	return (0);
4453 }
4454 
4455 static int
4456 em_sysctl_debug_info(SYSCTL_HANDLER_ARGS)
4457 {
4458 	struct adapter *adapter;
4459 	int error;
4460 	int result;
4461 
4462 	result = -1;
4463 	error = sysctl_handle_int(oidp, &result, 0, req);
4464 
4465 	if (error || !req->newptr)
4466 		return (error);
4467 
4468 	if (result == 1) {
4469 		adapter = (struct adapter *) arg1;
4470 		em_print_debug_info(adapter);
4471 	}
4472 
4473 	return (error);
4474 }
4475 
4476 static int
4477 em_get_rs(SYSCTL_HANDLER_ARGS)
4478 {
4479 	struct adapter *adapter = (struct adapter *) arg1;
4480 	int error;
4481 	int result;
4482 
4483 	result = 0;
4484 	error = sysctl_handle_int(oidp, &result, 0, req);
4485 
4486 	if (error || !req->newptr || result != 1)
4487 		return (error);
4488 	em_dump_rs(adapter);
4489 
4490 	return (error);
4491 }
4492 
4493 static void
4494 em_if_debug(if_ctx_t ctx)
4495 {
4496 	em_dump_rs(iflib_get_softc(ctx));
4497 }
4498 
4499 /*
4500  * This routine is meant to be fluid, add whatever is
4501  * needed for debugging a problem.  -jfv
4502  */
4503 static void
4504 em_print_debug_info(struct adapter *adapter)
4505 {
4506 	device_t dev = iflib_get_dev(adapter->ctx);
4507 	struct ifnet *ifp = iflib_get_ifp(adapter->ctx);
4508 	struct tx_ring *txr = &adapter->tx_queues->txr;
4509 	struct rx_ring *rxr = &adapter->rx_queues->rxr;
4510 
4511 	if (if_getdrvflags(ifp) & IFF_DRV_RUNNING)
4512 		printf("Interface is RUNNING ");
4513 	else
4514 		printf("Interface is NOT RUNNING\n");
4515 
4516 	if (if_getdrvflags(ifp) & IFF_DRV_OACTIVE)
4517 		printf("and INACTIVE\n");
4518 	else
4519 		printf("and ACTIVE\n");
4520 
4521 	for (int i = 0; i < adapter->tx_num_queues; i++, txr++) {
4522 		device_printf(dev, "TX Queue %d ------\n", i);
4523 		device_printf(dev, "hw tdh = %d, hw tdt = %d\n",
4524 			E1000_READ_REG(&adapter->hw, E1000_TDH(i)),
4525 			E1000_READ_REG(&adapter->hw, E1000_TDT(i)));
4526 
4527 	}
4528 	for (int j=0; j < adapter->rx_num_queues; j++, rxr++) {
4529 		device_printf(dev, "RX Queue %d ------\n", j);
4530 		device_printf(dev, "hw rdh = %d, hw rdt = %d\n",
4531 			E1000_READ_REG(&adapter->hw, E1000_RDH(j)),
4532 			E1000_READ_REG(&adapter->hw, E1000_RDT(j)));
4533 	}
4534 }
4535 
4536 /*
4537  * 82574 only:
4538  * Write a new value to the EEPROM increasing the number of MSIX
4539  * vectors from 3 to 5, for proper multiqueue support.
4540  */
4541 static void
4542 em_enable_vectors_82574(if_ctx_t ctx)
4543 {
4544 	struct adapter *adapter = iflib_get_softc(ctx);
4545 	struct e1000_hw *hw = &adapter->hw;
4546 	device_t dev = iflib_get_dev(ctx);
4547 	u16 edata;
4548 
4549 	e1000_read_nvm(hw, EM_NVM_PCIE_CTRL, 1, &edata);
4550 	printf("Current cap: %#06x\n", edata);
4551 	if (((edata & EM_NVM_MSIX_N_MASK) >> EM_NVM_MSIX_N_SHIFT) != 4) {
4552 		device_printf(dev, "Writing to eeprom: increasing "
4553 		    "reported MSIX vectors from 3 to 5...\n");
4554 		edata &= ~(EM_NVM_MSIX_N_MASK);
4555 		edata |= 4 << EM_NVM_MSIX_N_SHIFT;
4556 		e1000_write_nvm(hw, EM_NVM_PCIE_CTRL, 1, &edata);
4557 		e1000_update_nvm_checksum(hw);
4558 		device_printf(dev, "Writing to eeprom: done\n");
4559 	}
4560 }
4561