xref: /freebsd/sys/dev/e1000/if_em.c (revision d93a896e)
1 /*-
2  * Copyright (c) 2016 Matt Macy <mmacy@nextbsd.org>
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  *
14  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
15  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
16  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
17  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
18  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
19  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
20  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
22  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
23  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
24  * SUCH DAMAGE.
25  */
26 
27 /* $FreeBSD$ */
28 #include "if_em.h"
29 #include <sys/sbuf.h>
30 #include <machine/_inttypes.h>
31 
32 #define em_mac_min e1000_82547
33 #define igb_mac_min e1000_82575
34 
35 /*********************************************************************
36  *  Driver version:
37  *********************************************************************/
38 char em_driver_version[] = "7.6.1-k";
39 
40 /*********************************************************************
41  *  PCI Device ID Table
42  *
43  *  Used by probe to select devices to load on
44  *  Last field stores an index into e1000_strings
45  *  Last entry must be all 0s
46  *
47  *  { Vendor ID, Device ID, SubVendor ID, SubDevice ID, String Index }
48  *********************************************************************/
49 
50 static pci_vendor_info_t em_vendor_info_array[] =
51 {
52 	/* Intel(R) PRO/1000 Network Connection - Legacy em*/
53 	PVID(0x8086, E1000_DEV_ID_82540EM, "Intel(R) PRO/1000 Network Connection"),
54 	PVID(0x8086, E1000_DEV_ID_82540EM_LOM, "Intel(R) PRO/1000 Network Connection"),
55 	PVID(0x8086, E1000_DEV_ID_82540EP, "Intel(R) PRO/1000 Network Connection"),
56 	PVID(0x8086, E1000_DEV_ID_82540EP_LOM, "Intel(R) PRO/1000 Network Connection"),
57 	PVID(0x8086, E1000_DEV_ID_82540EP_LP, "Intel(R) PRO/1000 Network Connection"),
58 
59 	PVID(0x8086, E1000_DEV_ID_82541EI, "Intel(R) PRO/1000 Network Connection"),
60 	PVID(0x8086, E1000_DEV_ID_82541ER, "Intel(R) PRO/1000 Network Connection"),
61 	PVID(0x8086, E1000_DEV_ID_82541ER_LOM, "Intel(R) PRO/1000 Network Connection"),
62 	PVID(0x8086, E1000_DEV_ID_82541EI_MOBILE, "Intel(R) PRO/1000 Network Connection"),
63 	PVID(0x8086, E1000_DEV_ID_82541GI, "Intel(R) PRO/1000 Network Connection"),
64 	PVID(0x8086, E1000_DEV_ID_82541GI_LF, "Intel(R) PRO/1000 Network Connection"),
65 	PVID(0x8086, E1000_DEV_ID_82541GI_MOBILE, "Intel(R) PRO/1000 Network Connection"),
66 
67 	PVID(0x8086, E1000_DEV_ID_82542, "Intel(R) PRO/1000 Network Connection"),
68 
69 	PVID(0x8086, E1000_DEV_ID_82543GC_FIBER, "Intel(R) PRO/1000 Network Connection"),
70 	PVID(0x8086, E1000_DEV_ID_82543GC_COPPER, "Intel(R) PRO/1000 Network Connection"),
71 
72 	PVID(0x8086, E1000_DEV_ID_82544EI_COPPER, "Intel(R) PRO/1000 Network Connection"),
73 	PVID(0x8086, E1000_DEV_ID_82544EI_FIBER, "Intel(R) PRO/1000 Network Connection"),
74 	PVID(0x8086, E1000_DEV_ID_82544GC_COPPER, "Intel(R) PRO/1000 Network Connection"),
75 	PVID(0x8086, E1000_DEV_ID_82544GC_LOM, "Intel(R) PRO/1000 Network Connection"),
76 
77 	PVID(0x8086, E1000_DEV_ID_82545EM_COPPER, "Intel(R) PRO/1000 Network Connection"),
78 	PVID(0x8086, E1000_DEV_ID_82545EM_FIBER, "Intel(R) PRO/1000 Network Connection"),
79 	PVID(0x8086, E1000_DEV_ID_82545GM_COPPER, "Intel(R) PRO/1000 Network Connection"),
80 	PVID(0x8086, E1000_DEV_ID_82545GM_FIBER, "Intel(R) PRO/1000 Network Connection"),
81 	PVID(0x8086, E1000_DEV_ID_82545GM_SERDES, "Intel(R) PRO/1000 Network Connection"),
82 
83 	PVID(0x8086, E1000_DEV_ID_82546EB_COPPER, "Intel(R) PRO/1000 Network Connection"),
84 	PVID(0x8086, E1000_DEV_ID_82546EB_FIBER, "Intel(R) PRO/1000 Network Connection"),
85 	PVID(0x8086, E1000_DEV_ID_82546EB_QUAD_COPPER, "Intel(R) PRO/1000 Network Connection"),
86 	PVID(0x8086, E1000_DEV_ID_82546GB_COPPER, "Intel(R) PRO/1000 Network Connection"),
87 	PVID(0x8086, E1000_DEV_ID_82546GB_FIBER, "Intel(R) PRO/1000 Network Connection"),
88 	PVID(0x8086, E1000_DEV_ID_82546GB_SERDES, "Intel(R) PRO/1000 Network Connection"),
89 	PVID(0x8086, E1000_DEV_ID_82546GB_PCIE, "Intel(R) PRO/1000 Network Connection"),
90 	PVID(0x8086, E1000_DEV_ID_82546GB_QUAD_COPPER, "Intel(R) PRO/1000 Network Connection"),
91 	PVID(0x8086, E1000_DEV_ID_82546GB_QUAD_COPPER_KSP3, "Intel(R) PRO/1000 Network Connection"),
92 
93 	PVID(0x8086, E1000_DEV_ID_82547EI, "Intel(R) PRO/1000 Network Connection"),
94 	PVID(0x8086, E1000_DEV_ID_82547EI_MOBILE, "Intel(R) PRO/1000 Network Connection"),
95 	PVID(0x8086, E1000_DEV_ID_82547GI, "Intel(R) PRO/1000 Network Connection"),
96 
97 	/* Intel(R) PRO/1000 Network Connection - em */
98 	PVID(0x8086, E1000_DEV_ID_82571EB_COPPER, "Intel(R) PRO/1000 Network Connection"),
99 	PVID(0x8086, E1000_DEV_ID_82571EB_FIBER, "Intel(R) PRO/1000 Network Connection"),
100 	PVID(0x8086, E1000_DEV_ID_82571EB_SERDES, "Intel(R) PRO/1000 Network Connection"),
101 	PVID(0x8086, E1000_DEV_ID_82571EB_SERDES_DUAL, "Intel(R) PRO/1000 Network Connection"),
102 	PVID(0x8086, E1000_DEV_ID_82571EB_SERDES_QUAD, "Intel(R) PRO/1000 Network Connection"),
103 	PVID(0x8086, E1000_DEV_ID_82571EB_QUAD_COPPER, "Intel(R) PRO/1000 Network Connection"),
104 	PVID(0x8086, E1000_DEV_ID_82571EB_QUAD_COPPER_LP, "Intel(R) PRO/1000 Network Connection"),
105 	PVID(0x8086, E1000_DEV_ID_82571EB_QUAD_FIBER, "Intel(R) PRO/1000 Network Connection"),
106 	PVID(0x8086, E1000_DEV_ID_82571PT_QUAD_COPPER, "Intel(R) PRO/1000 Network Connection"),
107 	PVID(0x8086, E1000_DEV_ID_82572EI, "Intel(R) PRO/1000 Network Connection"),
108 	PVID(0x8086, E1000_DEV_ID_82572EI_COPPER, "Intel(R) PRO/1000 Network Connection"),
109 	PVID(0x8086, E1000_DEV_ID_82572EI_FIBER, "Intel(R) PRO/1000 Network Connection"),
110 	PVID(0x8086, E1000_DEV_ID_82572EI_SERDES, "Intel(R) PRO/1000 Network Connection"),
111 	PVID(0x8086, E1000_DEV_ID_82573E, "Intel(R) PRO/1000 Network Connection"),
112 	PVID(0x8086, E1000_DEV_ID_82573E_IAMT, "Intel(R) PRO/1000 Network Connection"),
113 	PVID(0x8086, E1000_DEV_ID_82573L, "Intel(R) PRO/1000 Network Connection"),
114 	PVID(0x8086, E1000_DEV_ID_82583V, "Intel(R) PRO/1000 Network Connection"),
115 	PVID(0x8086, E1000_DEV_ID_80003ES2LAN_COPPER_SPT, "Intel(R) PRO/1000 Network Connection"),
116 	PVID(0x8086, E1000_DEV_ID_80003ES2LAN_SERDES_SPT, "Intel(R) PRO/1000 Network Connection"),
117 	PVID(0x8086, E1000_DEV_ID_80003ES2LAN_COPPER_DPT, "Intel(R) PRO/1000 Network Connection"),
118 	PVID(0x8086, E1000_DEV_ID_80003ES2LAN_SERDES_DPT, "Intel(R) PRO/1000 Network Connection"),
119 	PVID(0x8086, E1000_DEV_ID_ICH8_IGP_M_AMT, "Intel(R) PRO/1000 Network Connection"),
120 	PVID(0x8086, E1000_DEV_ID_ICH8_IGP_AMT, "Intel(R) PRO/1000 Network Connection"),
121 	PVID(0x8086, E1000_DEV_ID_ICH8_IGP_C, "Intel(R) PRO/1000 Network Connection"),
122 	PVID(0x8086, E1000_DEV_ID_ICH8_IFE, "Intel(R) PRO/1000 Network Connection"),
123 	PVID(0x8086, E1000_DEV_ID_ICH8_IFE_GT, "Intel(R) PRO/1000 Network Connection"),
124 	PVID(0x8086, E1000_DEV_ID_ICH8_IFE_G, "Intel(R) PRO/1000 Network Connection"),
125 	PVID(0x8086, E1000_DEV_ID_ICH8_IGP_M, "Intel(R) PRO/1000 Network Connection"),
126 	PVID(0x8086, E1000_DEV_ID_ICH8_82567V_3, "Intel(R) PRO/1000 Network Connection"),
127 	PVID(0x8086, E1000_DEV_ID_ICH9_IGP_M_AMT, "Intel(R) PRO/1000 Network Connection"),
128 	PVID(0x8086, E1000_DEV_ID_ICH9_IGP_AMT, "Intel(R) PRO/1000 Network Connection"),
129 	PVID(0x8086, E1000_DEV_ID_ICH9_IGP_C, "Intel(R) PRO/1000 Network Connection"),
130 	PVID(0x8086, E1000_DEV_ID_ICH9_IGP_M, "Intel(R) PRO/1000 Network Connection"),
131 	PVID(0x8086, E1000_DEV_ID_ICH9_IGP_M_V, "Intel(R) PRO/1000 Network Connection"),
132 	PVID(0x8086, E1000_DEV_ID_ICH9_IFE, "Intel(R) PRO/1000 Network Connection"),
133 	PVID(0x8086, E1000_DEV_ID_ICH9_IFE_GT, "Intel(R) PRO/1000 Network Connection"),
134 	PVID(0x8086, E1000_DEV_ID_ICH9_IFE_G, "Intel(R) PRO/1000 Network Connection"),
135 	PVID(0x8086, E1000_DEV_ID_ICH9_BM, "Intel(R) PRO/1000 Network Connection"),
136 	PVID(0x8086, E1000_DEV_ID_82574L, "Intel(R) PRO/1000 Network Connection"),
137 	PVID(0x8086, E1000_DEV_ID_82574LA, "Intel(R) PRO/1000 Network Connection"),
138 	PVID(0x8086, E1000_DEV_ID_ICH10_R_BM_LM, "Intel(R) PRO/1000 Network Connection"),
139 	PVID(0x8086, E1000_DEV_ID_ICH10_R_BM_LF, "Intel(R) PRO/1000 Network Connection"),
140 	PVID(0x8086, E1000_DEV_ID_ICH10_R_BM_V, "Intel(R) PRO/1000 Network Connection"),
141 	PVID(0x8086, E1000_DEV_ID_ICH10_D_BM_LM, "Intel(R) PRO/1000 Network Connection"),
142 	PVID(0x8086, E1000_DEV_ID_ICH10_D_BM_LF, "Intel(R) PRO/1000 Network Connection"),
143 	PVID(0x8086, E1000_DEV_ID_ICH10_D_BM_V, "Intel(R) PRO/1000 Network Connection"),
144 	PVID(0x8086, E1000_DEV_ID_PCH_M_HV_LM, "Intel(R) PRO/1000 Network Connection"),
145 	PVID(0x8086, E1000_DEV_ID_PCH_M_HV_LC, "Intel(R) PRO/1000 Network Connection"),
146 	PVID(0x8086, E1000_DEV_ID_PCH_D_HV_DM, "Intel(R) PRO/1000 Network Connection"),
147 	PVID(0x8086, E1000_DEV_ID_PCH_D_HV_DC, "Intel(R) PRO/1000 Network Connection"),
148 	PVID(0x8086, E1000_DEV_ID_PCH2_LV_LM, "Intel(R) PRO/1000 Network Connection"),
149 	PVID(0x8086, E1000_DEV_ID_PCH2_LV_V, "Intel(R) PRO/1000 Network Connection"),
150 	PVID(0x8086, E1000_DEV_ID_PCH_LPT_I217_LM, "Intel(R) PRO/1000 Network Connection"),
151 	PVID(0x8086, E1000_DEV_ID_PCH_LPT_I217_V, "Intel(R) PRO/1000 Network Connection"),
152 	PVID(0x8086, E1000_DEV_ID_PCH_LPTLP_I218_LM, "Intel(R) PRO/1000 Network Connection"),
153 	PVID(0x8086, E1000_DEV_ID_PCH_LPTLP_I218_V, "Intel(R) PRO/1000 Network Connection"),
154 	PVID(0x8086, E1000_DEV_ID_PCH_I218_LM2, "Intel(R) PRO/1000 Network Connection"),
155 	PVID(0x8086, E1000_DEV_ID_PCH_I218_V2, "Intel(R) PRO/1000 Network Connection"),
156 	PVID(0x8086, E1000_DEV_ID_PCH_I218_LM3, "Intel(R) PRO/1000 Network Connection"),
157 	PVID(0x8086, E1000_DEV_ID_PCH_I218_V3, "Intel(R) PRO/1000 Network Connection"),
158 	PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_LM, "Intel(R) PRO/1000 Network Connection"),
159 	PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_V, "Intel(R) PRO/1000 Network Connection"),
160 	PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_LM2, "Intel(R) PRO/1000 Network Connection"),
161 	PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_V2, "Intel(R) PRO/1000 Network Connection"),
162 	PVID(0x8086, E1000_DEV_ID_PCH_LBG_I219_LM3, "Intel(R) PRO/1000 Network Connection"),
163 	PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_LM4, "Intel(R) PRO/1000 Network Connection"),
164 	PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_V4, "Intel(R) PRO/1000 Network Connection"),
165 	PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_LM5, "Intel(R) PRO/1000 Network Connection"),
166 	PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_V5, "Intel(R) PRO/1000 Network Connection"),
167 	/* required last entry */
168 	PVID_END
169 };
170 
171 static pci_vendor_info_t igb_vendor_info_array[] =
172 {
173 	/* Intel(R) PRO/1000 Network Connection - igb */
174 	PVID(0x8086, E1000_DEV_ID_82575EB_COPPER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
175 	PVID(0x8086, E1000_DEV_ID_82575EB_FIBER_SERDES, "Intel(R) PRO/1000 PCI-Express Network Driver"),
176 	PVID(0x8086, E1000_DEV_ID_82575GB_QUAD_COPPER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
177 	PVID(0x8086, E1000_DEV_ID_82576, "Intel(R) PRO/1000 PCI-Express Network Driver"),
178 	PVID(0x8086, E1000_DEV_ID_82576_NS, "Intel(R) PRO/1000 PCI-Express Network Driver"),
179 	PVID(0x8086, E1000_DEV_ID_82576_NS_SERDES, "Intel(R) PRO/1000 PCI-Express Network Driver"),
180 	PVID(0x8086, E1000_DEV_ID_82576_FIBER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
181 	PVID(0x8086, E1000_DEV_ID_82576_SERDES, "Intel(R) PRO/1000 PCI-Express Network Driver"),
182 	PVID(0x8086, E1000_DEV_ID_82576_SERDES_QUAD, "Intel(R) PRO/1000 PCI-Express Network Driver"),
183 	PVID(0x8086, E1000_DEV_ID_82576_QUAD_COPPER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
184 	PVID(0x8086, E1000_DEV_ID_82576_QUAD_COPPER_ET2, "Intel(R) PRO/1000 PCI-Express Network Driver"),
185 	PVID(0x8086, E1000_DEV_ID_82576_VF, "Intel(R) PRO/1000 PCI-Express Network Driver"),
186 	PVID(0x8086, E1000_DEV_ID_82580_COPPER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
187 	PVID(0x8086, E1000_DEV_ID_82580_FIBER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
188 	PVID(0x8086, E1000_DEV_ID_82580_SERDES, "Intel(R) PRO/1000 PCI-Express Network Driver"),
189 	PVID(0x8086, E1000_DEV_ID_82580_SGMII, "Intel(R) PRO/1000 PCI-Express Network Driver"),
190 	PVID(0x8086, E1000_DEV_ID_82580_COPPER_DUAL, "Intel(R) PRO/1000 PCI-Express Network Driver"),
191 	PVID(0x8086, E1000_DEV_ID_82580_QUAD_FIBER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
192 	PVID(0x8086, E1000_DEV_ID_DH89XXCC_SERDES, "Intel(R) PRO/1000 PCI-Express Network Driver"),
193 	PVID(0x8086, E1000_DEV_ID_DH89XXCC_SGMII, "Intel(R) PRO/1000 PCI-Express Network Driver"),
194 	PVID(0x8086, E1000_DEV_ID_DH89XXCC_SFP, "Intel(R) PRO/1000 PCI-Express Network Driver"),
195 	PVID(0x8086, E1000_DEV_ID_DH89XXCC_BACKPLANE, "Intel(R) PRO/1000 PCI-Express Network Driver"),
196 	PVID(0x8086, E1000_DEV_ID_I350_COPPER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
197 	PVID(0x8086, E1000_DEV_ID_I350_FIBER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
198 	PVID(0x8086, E1000_DEV_ID_I350_SERDES, "Intel(R) PRO/1000 PCI-Express Network Driver"),
199 	PVID(0x8086, E1000_DEV_ID_I350_SGMII, "Intel(R) PRO/1000 PCI-Express Network Driver"),
200 	PVID(0x8086, E1000_DEV_ID_I350_VF, "Intel(R) PRO/1000 PCI-Express Network Driver"),
201 	PVID(0x8086, E1000_DEV_ID_I210_COPPER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
202 	PVID(0x8086, E1000_DEV_ID_I210_COPPER_IT, "Intel(R) PRO/1000 PCI-Express Network Driver"),
203 	PVID(0x8086, E1000_DEV_ID_I210_COPPER_OEM1, "Intel(R) PRO/1000 PCI-Express Network Driver"),
204 	PVID(0x8086, E1000_DEV_ID_I210_COPPER_FLASHLESS, "Intel(R) PRO/1000 PCI-Express Network Driver"),
205 	PVID(0x8086, E1000_DEV_ID_I210_SERDES_FLASHLESS, "Intel(R) PRO/1000 PCI-Express Network Driver"),
206 	PVID(0x8086, E1000_DEV_ID_I210_FIBER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
207 	PVID(0x8086, E1000_DEV_ID_I210_SERDES, "Intel(R) PRO/1000 PCI-Express Network Driver"),
208 	PVID(0x8086, E1000_DEV_ID_I210_SGMII, "Intel(R) PRO/1000 PCI-Express Network Driver"),
209 	PVID(0x8086, E1000_DEV_ID_I211_COPPER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
210 	PVID(0x8086, E1000_DEV_ID_I354_BACKPLANE_1GBPS, "Intel(R) PRO/1000 PCI-Express Network Driver"),
211 	PVID(0x8086, E1000_DEV_ID_I354_BACKPLANE_2_5GBPS, "Intel(R) PRO/1000 PCI-Express Network Driver"),
212 	PVID(0x8086, E1000_DEV_ID_I354_SGMII, "Intel(R) PRO/1000 PCI-Express Network Driver"),
213 	/* required last entry */
214 	PVID_END
215 };
216 
217 /*********************************************************************
218  *  Function prototypes
219  *********************************************************************/
220 static void	*em_register(device_t dev);
221 static void	*igb_register(device_t dev);
222 static int	em_if_attach_pre(if_ctx_t ctx);
223 static int	em_if_attach_post(if_ctx_t ctx);
224 static int	em_if_detach(if_ctx_t ctx);
225 static int	em_if_shutdown(if_ctx_t ctx);
226 static int	em_if_suspend(if_ctx_t ctx);
227 static int	em_if_resume(if_ctx_t ctx);
228 
229 static int	em_if_tx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int ntxqs, int ntxqsets);
230 static int	em_if_rx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int nrxqs, int nrxqsets);
231 static void	em_if_queues_free(if_ctx_t ctx);
232 
233 static uint64_t	em_if_get_counter(if_ctx_t, ift_counter);
234 static void	em_if_init(if_ctx_t ctx);
235 static void	em_if_stop(if_ctx_t ctx);
236 static void	em_if_media_status(if_ctx_t, struct ifmediareq *);
237 static int	em_if_media_change(if_ctx_t ctx);
238 static int	em_if_mtu_set(if_ctx_t ctx, uint32_t mtu);
239 static void	em_if_timer(if_ctx_t ctx, uint16_t qid);
240 static void	em_if_vlan_register(if_ctx_t ctx, u16 vtag);
241 static void	em_if_vlan_unregister(if_ctx_t ctx, u16 vtag);
242 
243 static void	em_identify_hardware(if_ctx_t ctx);
244 static int	em_allocate_pci_resources(if_ctx_t ctx);
245 static void	em_free_pci_resources(if_ctx_t ctx);
246 static void	em_reset(if_ctx_t ctx);
247 static int	em_setup_interface(if_ctx_t ctx);
248 static int	em_setup_msix(if_ctx_t ctx);
249 
250 static void	em_initialize_transmit_unit(if_ctx_t ctx);
251 static void	em_initialize_receive_unit(if_ctx_t ctx);
252 
253 static void	em_if_enable_intr(if_ctx_t ctx);
254 static void	em_if_disable_intr(if_ctx_t ctx);
255 static int	em_if_rx_queue_intr_enable(if_ctx_t ctx, uint16_t rxqid);
256 static int	em_if_tx_queue_intr_enable(if_ctx_t ctx, uint16_t txqid);
257 static void	em_if_multi_set(if_ctx_t ctx);
258 static void	em_if_update_admin_status(if_ctx_t ctx);
259 static void	em_if_debug(if_ctx_t ctx);
260 static void	em_update_stats_counters(struct adapter *);
261 static void	em_add_hw_stats(struct adapter *adapter);
262 static int	em_if_set_promisc(if_ctx_t ctx, int flags);
263 static void	em_setup_vlan_hw_support(struct adapter *);
264 static int	em_sysctl_nvm_info(SYSCTL_HANDLER_ARGS);
265 static void	em_print_nvm_info(struct adapter *);
266 static int	em_sysctl_debug_info(SYSCTL_HANDLER_ARGS);
267 static int	em_get_rs(SYSCTL_HANDLER_ARGS);
268 static void	em_print_debug_info(struct adapter *);
269 static int 	em_is_valid_ether_addr(u8 *);
270 static int	em_sysctl_int_delay(SYSCTL_HANDLER_ARGS);
271 static void	em_add_int_delay_sysctl(struct adapter *, const char *,
272 		    const char *, struct em_int_delay_info *, int, int);
273 /* Management and WOL Support */
274 static void	em_init_manageability(struct adapter *);
275 static void	em_release_manageability(struct adapter *);
276 static void	em_get_hw_control(struct adapter *);
277 static void	em_release_hw_control(struct adapter *);
278 static void	em_get_wakeup(if_ctx_t ctx);
279 static void	em_enable_wakeup(if_ctx_t ctx);
280 static int	em_enable_phy_wakeup(struct adapter *);
281 static void	em_disable_aspm(struct adapter *);
282 
283 int		em_intr(void *arg);
284 static void	em_disable_promisc(if_ctx_t ctx);
285 
286 /* MSIX handlers */
287 static int	em_if_msix_intr_assign(if_ctx_t, int);
288 static int	em_msix_link(void *);
289 static void	em_handle_link(void *context);
290 
291 static void	em_enable_vectors_82574(if_ctx_t);
292 
293 static int	em_set_flowcntl(SYSCTL_HANDLER_ARGS);
294 static int	em_sysctl_eee(SYSCTL_HANDLER_ARGS);
295 static void	em_if_led_func(if_ctx_t ctx, int onoff);
296 
297 static int	em_get_regs(SYSCTL_HANDLER_ARGS);
298 
299 static void	lem_smartspeed(struct adapter *adapter);
300 static void	igb_configure_queues(struct adapter *adapter);
301 
302 
303 /*********************************************************************
304  *  FreeBSD Device Interface Entry Points
305  *********************************************************************/
306 static device_method_t em_methods[] = {
307 	/* Device interface */
308 	DEVMETHOD(device_register, em_register),
309 	DEVMETHOD(device_probe, iflib_device_probe),
310 	DEVMETHOD(device_attach, iflib_device_attach),
311 	DEVMETHOD(device_detach, iflib_device_detach),
312 	DEVMETHOD(device_shutdown, iflib_device_shutdown),
313 	DEVMETHOD(device_suspend, iflib_device_suspend),
314 	DEVMETHOD(device_resume, iflib_device_resume),
315 	DEVMETHOD_END
316 };
317 
318 static device_method_t igb_methods[] = {
319 	/* Device interface */
320 	DEVMETHOD(device_register, igb_register),
321 	DEVMETHOD(device_probe, iflib_device_probe),
322 	DEVMETHOD(device_attach, iflib_device_attach),
323 	DEVMETHOD(device_detach, iflib_device_detach),
324 	DEVMETHOD(device_shutdown, iflib_device_shutdown),
325 	DEVMETHOD(device_suspend, iflib_device_suspend),
326 	DEVMETHOD(device_resume, iflib_device_resume),
327 	DEVMETHOD_END
328 };
329 
330 
331 static driver_t em_driver = {
332 	"em", em_methods, sizeof(struct adapter),
333 };
334 
335 static devclass_t em_devclass;
336 DRIVER_MODULE(em, pci, em_driver, em_devclass, 0, 0);
337 
338 MODULE_DEPEND(em, pci, 1, 1, 1);
339 MODULE_DEPEND(em, ether, 1, 1, 1);
340 MODULE_DEPEND(em, iflib, 1, 1, 1);
341 
342 static driver_t igb_driver = {
343 	"igb", igb_methods, sizeof(struct adapter),
344 };
345 
346 static devclass_t igb_devclass;
347 DRIVER_MODULE(igb, pci, igb_driver, igb_devclass, 0, 0);
348 
349 MODULE_DEPEND(igb, pci, 1, 1, 1);
350 MODULE_DEPEND(igb, ether, 1, 1, 1);
351 MODULE_DEPEND(igb, iflib, 1, 1, 1);
352 
353 
354 static device_method_t em_if_methods[] = {
355 	DEVMETHOD(ifdi_attach_pre, em_if_attach_pre),
356 	DEVMETHOD(ifdi_attach_post, em_if_attach_post),
357 	DEVMETHOD(ifdi_detach, em_if_detach),
358 	DEVMETHOD(ifdi_shutdown, em_if_shutdown),
359 	DEVMETHOD(ifdi_suspend, em_if_suspend),
360 	DEVMETHOD(ifdi_resume, em_if_resume),
361 	DEVMETHOD(ifdi_init, em_if_init),
362 	DEVMETHOD(ifdi_stop, em_if_stop),
363 	DEVMETHOD(ifdi_msix_intr_assign, em_if_msix_intr_assign),
364 	DEVMETHOD(ifdi_intr_enable, em_if_enable_intr),
365 	DEVMETHOD(ifdi_intr_disable, em_if_disable_intr),
366 	DEVMETHOD(ifdi_tx_queues_alloc, em_if_tx_queues_alloc),
367 	DEVMETHOD(ifdi_rx_queues_alloc, em_if_rx_queues_alloc),
368 	DEVMETHOD(ifdi_queues_free, em_if_queues_free),
369 	DEVMETHOD(ifdi_update_admin_status, em_if_update_admin_status),
370 	DEVMETHOD(ifdi_multi_set, em_if_multi_set),
371 	DEVMETHOD(ifdi_media_status, em_if_media_status),
372 	DEVMETHOD(ifdi_media_change, em_if_media_change),
373 	DEVMETHOD(ifdi_mtu_set, em_if_mtu_set),
374 	DEVMETHOD(ifdi_promisc_set, em_if_set_promisc),
375 	DEVMETHOD(ifdi_timer, em_if_timer),
376 	DEVMETHOD(ifdi_vlan_register, em_if_vlan_register),
377 	DEVMETHOD(ifdi_vlan_unregister, em_if_vlan_unregister),
378 	DEVMETHOD(ifdi_get_counter, em_if_get_counter),
379 	DEVMETHOD(ifdi_led_func, em_if_led_func),
380 	DEVMETHOD(ifdi_rx_queue_intr_enable, em_if_rx_queue_intr_enable),
381 	DEVMETHOD(ifdi_tx_queue_intr_enable, em_if_tx_queue_intr_enable),
382 	DEVMETHOD(ifdi_debug, em_if_debug),
383 	DEVMETHOD_END
384 };
385 
386 /*
387  * note that if (adapter->msix_mem) is replaced by:
388  * if (adapter->intr_type == IFLIB_INTR_MSIX)
389  */
390 static driver_t em_if_driver = {
391 	"em_if", em_if_methods, sizeof(struct adapter)
392 };
393 
394 /*********************************************************************
395  *  Tunable default values.
396  *********************************************************************/
397 
398 #define EM_TICKS_TO_USECS(ticks)	((1024 * (ticks) + 500) / 1000)
399 #define EM_USECS_TO_TICKS(usecs)	((1000 * (usecs) + 512) / 1024)
400 #define M_TSO_LEN			66
401 
402 #define MAX_INTS_PER_SEC	8000
403 #define DEFAULT_ITR		(1000000000/(MAX_INTS_PER_SEC * 256))
404 
405 /* Allow common code without TSO */
406 #ifndef CSUM_TSO
407 #define CSUM_TSO	0
408 #endif
409 
410 #define TSO_WORKAROUND	4
411 
412 static SYSCTL_NODE(_hw, OID_AUTO, em, CTLFLAG_RD, 0, "EM driver parameters");
413 
414 static int em_disable_crc_stripping = 0;
415 SYSCTL_INT(_hw_em, OID_AUTO, disable_crc_stripping, CTLFLAG_RDTUN,
416     &em_disable_crc_stripping, 0, "Disable CRC Stripping");
417 
418 static int em_tx_int_delay_dflt = EM_TICKS_TO_USECS(EM_TIDV);
419 static int em_rx_int_delay_dflt = EM_TICKS_TO_USECS(EM_RDTR);
420 SYSCTL_INT(_hw_em, OID_AUTO, tx_int_delay, CTLFLAG_RDTUN, &em_tx_int_delay_dflt,
421     0, "Default transmit interrupt delay in usecs");
422 SYSCTL_INT(_hw_em, OID_AUTO, rx_int_delay, CTLFLAG_RDTUN, &em_rx_int_delay_dflt,
423     0, "Default receive interrupt delay in usecs");
424 
425 static int em_tx_abs_int_delay_dflt = EM_TICKS_TO_USECS(EM_TADV);
426 static int em_rx_abs_int_delay_dflt = EM_TICKS_TO_USECS(EM_RADV);
427 SYSCTL_INT(_hw_em, OID_AUTO, tx_abs_int_delay, CTLFLAG_RDTUN,
428     &em_tx_abs_int_delay_dflt, 0,
429     "Default transmit interrupt delay limit in usecs");
430 SYSCTL_INT(_hw_em, OID_AUTO, rx_abs_int_delay, CTLFLAG_RDTUN,
431     &em_rx_abs_int_delay_dflt, 0,
432     "Default receive interrupt delay limit in usecs");
433 
434 static int em_smart_pwr_down = FALSE;
435 SYSCTL_INT(_hw_em, OID_AUTO, smart_pwr_down, CTLFLAG_RDTUN, &em_smart_pwr_down,
436     0, "Set to true to leave smart power down enabled on newer adapters");
437 
438 /* Controls whether promiscuous also shows bad packets */
439 static int em_debug_sbp = TRUE;
440 SYSCTL_INT(_hw_em, OID_AUTO, sbp, CTLFLAG_RDTUN, &em_debug_sbp, 0,
441     "Show bad packets in promiscuous mode");
442 
443 /* How many packets rxeof tries to clean at a time */
444 static int em_rx_process_limit = 100;
445 SYSCTL_INT(_hw_em, OID_AUTO, rx_process_limit, CTLFLAG_RDTUN,
446     &em_rx_process_limit, 0,
447     "Maximum number of received packets to process "
448     "at a time, -1 means unlimited");
449 
450 /* Energy efficient ethernet - default to OFF */
451 static int eee_setting = 1;
452 SYSCTL_INT(_hw_em, OID_AUTO, eee_setting, CTLFLAG_RDTUN, &eee_setting, 0,
453     "Enable Energy Efficient Ethernet");
454 
455 /*
456 ** Tuneable Interrupt rate
457 */
458 static int em_max_interrupt_rate = 8000;
459 SYSCTL_INT(_hw_em, OID_AUTO, max_interrupt_rate, CTLFLAG_RDTUN,
460     &em_max_interrupt_rate, 0, "Maximum interrupts per second");
461 
462 
463 
464 /* Global used in WOL setup with multiport cards */
465 static int global_quad_port_a = 0;
466 
467 extern struct if_txrx igb_txrx;
468 extern struct if_txrx em_txrx;
469 extern struct if_txrx lem_txrx;
470 
471 static struct if_shared_ctx em_sctx_init = {
472 	.isc_magic = IFLIB_MAGIC,
473 	.isc_q_align = PAGE_SIZE,
474 	.isc_tx_maxsize = EM_TSO_SIZE,
475 	.isc_tx_maxsegsize = PAGE_SIZE,
476 	.isc_rx_maxsize = MJUM9BYTES,
477 	.isc_rx_nsegments = 1,
478 	.isc_rx_maxsegsize = MJUM9BYTES,
479 	.isc_nfl = 1,
480 	.isc_nrxqs = 1,
481 	.isc_ntxqs = 1,
482 	.isc_admin_intrcnt = 1,
483 	.isc_vendor_info = em_vendor_info_array,
484 	.isc_driver_version = em_driver_version,
485 	.isc_driver = &em_if_driver,
486 	.isc_flags = IFLIB_NEED_SCRATCH | IFLIB_TSO_INIT_IP,
487 
488 	.isc_nrxd_min = {EM_MIN_RXD},
489 	.isc_ntxd_min = {EM_MIN_TXD},
490 	.isc_nrxd_max = {EM_MAX_RXD},
491 	.isc_ntxd_max = {EM_MAX_TXD},
492 	.isc_nrxd_default = {EM_DEFAULT_RXD},
493 	.isc_ntxd_default = {EM_DEFAULT_TXD},
494 };
495 
496 if_shared_ctx_t em_sctx = &em_sctx_init;
497 
498 
499 static struct if_shared_ctx igb_sctx_init = {
500 	.isc_magic = IFLIB_MAGIC,
501 	.isc_q_align = PAGE_SIZE,
502 	.isc_tx_maxsize = EM_TSO_SIZE,
503 	.isc_tx_maxsegsize = PAGE_SIZE,
504 	.isc_rx_maxsize = MJUM9BYTES,
505 	.isc_rx_nsegments = 1,
506 	.isc_rx_maxsegsize = MJUM9BYTES,
507 	.isc_nfl = 1,
508 	.isc_nrxqs = 1,
509 	.isc_ntxqs = 1,
510 	.isc_admin_intrcnt = 1,
511 	.isc_vendor_info = igb_vendor_info_array,
512 	.isc_driver_version = em_driver_version,
513 	.isc_driver = &em_if_driver,
514 	.isc_flags = IFLIB_NEED_SCRATCH | IFLIB_TSO_INIT_IP,
515 
516 	.isc_nrxd_min = {EM_MIN_RXD},
517 	.isc_ntxd_min = {EM_MIN_TXD},
518 	.isc_nrxd_max = {EM_MAX_RXD},
519 	.isc_ntxd_max = {EM_MAX_TXD},
520 	.isc_nrxd_default = {EM_DEFAULT_RXD},
521 	.isc_ntxd_default = {EM_DEFAULT_TXD},
522 };
523 
524 if_shared_ctx_t igb_sctx = &igb_sctx_init;
525 
526 /*****************************************************************
527  *
528  * Dump Registers
529  *
530  ****************************************************************/
531 #define IGB_REGS_LEN 739
532 
533 static int em_get_regs(SYSCTL_HANDLER_ARGS)
534 {
535 	struct adapter *adapter = (struct adapter *)arg1;
536 	struct e1000_hw *hw = &adapter->hw;
537 
538 	struct sbuf *sb;
539 	u32 *regs_buff = (u32 *)malloc(sizeof(u32) * IGB_REGS_LEN, M_DEVBUF, M_NOWAIT);
540 	int rc;
541 
542 	memset(regs_buff, 0, IGB_REGS_LEN * sizeof(u32));
543 
544 	rc = sysctl_wire_old_buffer(req, 0);
545 	MPASS(rc == 0);
546 	if (rc != 0)
547 		return (rc);
548 
549 	sb = sbuf_new_for_sysctl(NULL, NULL, 32*400, req);
550 	MPASS(sb != NULL);
551 	if (sb == NULL)
552 		return (ENOMEM);
553 
554 	/* General Registers */
555 	regs_buff[0] = E1000_READ_REG(hw, E1000_CTRL);
556 	regs_buff[1] = E1000_READ_REG(hw, E1000_STATUS);
557 	regs_buff[2] = E1000_READ_REG(hw, E1000_CTRL_EXT);
558 	regs_buff[3] = E1000_READ_REG(hw, E1000_ICR);
559 	regs_buff[4] = E1000_READ_REG(hw, E1000_RCTL);
560 	regs_buff[5] = E1000_READ_REG(hw, E1000_RDLEN(0));
561 	regs_buff[6] = E1000_READ_REG(hw, E1000_RDH(0));
562 	regs_buff[7] = E1000_READ_REG(hw, E1000_RDT(0));
563 	regs_buff[8] = E1000_READ_REG(hw, E1000_RXDCTL(0));
564 	regs_buff[9] = E1000_READ_REG(hw, E1000_RDBAL(0));
565 	regs_buff[10] = E1000_READ_REG(hw, E1000_RDBAH(0));
566 	regs_buff[11] = E1000_READ_REG(hw, E1000_TCTL);
567 	regs_buff[12] = E1000_READ_REG(hw, E1000_TDBAL(0));
568 	regs_buff[13] = E1000_READ_REG(hw, E1000_TDBAH(0));
569 	regs_buff[14] = E1000_READ_REG(hw, E1000_TDLEN(0));
570 	regs_buff[15] = E1000_READ_REG(hw, E1000_TDH(0));
571 	regs_buff[16] = E1000_READ_REG(hw, E1000_TDT(0));
572 	regs_buff[17] = E1000_READ_REG(hw, E1000_TXDCTL(0));
573 	regs_buff[18] = E1000_READ_REG(hw, E1000_TDFH);
574 	regs_buff[19] = E1000_READ_REG(hw, E1000_TDFT);
575 	regs_buff[20] = E1000_READ_REG(hw, E1000_TDFHS);
576 	regs_buff[21] = E1000_READ_REG(hw, E1000_TDFPC);
577 
578 	sbuf_printf(sb, "General Registers\n");
579 	sbuf_printf(sb, "\tCTRL\t %08x\n", regs_buff[0]);
580 	sbuf_printf(sb, "\tSTATUS\t %08x\n", regs_buff[1]);
581 	sbuf_printf(sb, "\tCTRL_EXIT\t %08x\n\n", regs_buff[2]);
582 
583 	sbuf_printf(sb, "Interrupt Registers\n");
584 	sbuf_printf(sb, "\tICR\t %08x\n\n", regs_buff[3]);
585 
586 	sbuf_printf(sb, "RX Registers\n");
587 	sbuf_printf(sb, "\tRCTL\t %08x\n", regs_buff[4]);
588 	sbuf_printf(sb, "\tRDLEN\t %08x\n", regs_buff[5]);
589 	sbuf_printf(sb, "\tRDH\t %08x\n", regs_buff[6]);
590 	sbuf_printf(sb, "\tRDT\t %08x\n", regs_buff[7]);
591 	sbuf_printf(sb, "\tRXDCTL\t %08x\n", regs_buff[8]);
592 	sbuf_printf(sb, "\tRDBAL\t %08x\n", regs_buff[9]);
593 	sbuf_printf(sb, "\tRDBAH\t %08x\n\n", regs_buff[10]);
594 
595 	sbuf_printf(sb, "TX Registers\n");
596 	sbuf_printf(sb, "\tTCTL\t %08x\n", regs_buff[11]);
597 	sbuf_printf(sb, "\tTDBAL\t %08x\n", regs_buff[12]);
598 	sbuf_printf(sb, "\tTDBAH\t %08x\n", regs_buff[13]);
599 	sbuf_printf(sb, "\tTDLEN\t %08x\n", regs_buff[14]);
600 	sbuf_printf(sb, "\tTDH\t %08x\n", regs_buff[15]);
601 	sbuf_printf(sb, "\tTDT\t %08x\n", regs_buff[16]);
602 	sbuf_printf(sb, "\tTXDCTL\t %08x\n", regs_buff[17]);
603 	sbuf_printf(sb, "\tTDFH\t %08x\n", regs_buff[18]);
604 	sbuf_printf(sb, "\tTDFT\t %08x\n", regs_buff[19]);
605 	sbuf_printf(sb, "\tTDFHS\t %08x\n", regs_buff[20]);
606 	sbuf_printf(sb, "\tTDFPC\t %08x\n\n", regs_buff[21]);
607 
608 #ifdef DUMP_DESCS
609 	{
610 		if_softc_ctx_t scctx = adapter->shared;
611 		struct rx_ring *rxr = &rx_que->rxr;
612 		struct tx_ring *txr = &tx_que->txr;
613 		int ntxd = scctx->isc_ntxd[0];
614 		int nrxd = scctx->isc_nrxd[0];
615 		int j;
616 
617 	for (j = 0; j < nrxd; j++) {
618 		u32 staterr = le32toh(rxr->rx_base[j].wb.upper.status_error);
619 		u32 length =  le32toh(rxr->rx_base[j].wb.upper.length);
620 		sbuf_printf(sb, "\tReceive Descriptor Address %d: %08" PRIx64 "  Error:%d  Length:%d\n", j, rxr->rx_base[j].read.buffer_addr, staterr, length);
621 	}
622 
623 	for (j = 0; j < min(ntxd, 256); j++) {
624 		unsigned int *ptr = (unsigned int *)&txr->tx_base[j];
625 
626 		sbuf_printf(sb, "\tTXD[%03d] [0]: %08x [1]: %08x [2]: %08x [3]: %08x  eop: %d DD=%d\n",
627 			    j, ptr[0], ptr[1], ptr[2], ptr[3], buf->eop,
628 			    buf->eop != -1 ? txr->tx_base[buf->eop].upper.fields.status & E1000_TXD_STAT_DD : 0);
629 
630 	}
631 	}
632 #endif
633 
634 	rc = sbuf_finish(sb);
635 	sbuf_delete(sb);
636 	return(rc);
637 }
638 
639 static void *
640 em_register(device_t dev)
641 {
642 	return (em_sctx);
643 }
644 
645 static void *
646 igb_register(device_t dev)
647 {
648 	return (igb_sctx);
649 }
650 
651 static int
652 em_set_num_queues(if_ctx_t ctx)
653 {
654 	struct adapter *adapter = iflib_get_softc(ctx);
655 	int maxqueues;
656 
657 	/* Sanity check based on HW */
658 	switch (adapter->hw.mac.type) {
659 	case e1000_82576:
660 	case e1000_82580:
661 	case e1000_i350:
662 	case e1000_i354:
663 		maxqueues = 8;
664 		break;
665 	case e1000_i210:
666 	case e1000_82575:
667 		maxqueues = 4;
668 		break;
669 	case e1000_i211:
670 	case e1000_82574:
671 		maxqueues = 2;
672 		break;
673 	default:
674 		maxqueues = 1;
675 		break;
676 	}
677 
678 	return (maxqueues);
679 }
680 
681 
682 #define EM_CAPS \
683 	IFCAP_TSO4 | IFCAP_TXCSUM | IFCAP_LRO | IFCAP_RXCSUM | IFCAP_VLAN_HWFILTER | IFCAP_WOL_MAGIC | \
684 	IFCAP_WOL_MCAST | IFCAP_WOL | IFCAP_VLAN_HWTSO | IFCAP_HWCSUM | IFCAP_VLAN_HWTAGGING | \
685 	IFCAP_VLAN_HWCSUM | IFCAP_VLAN_HWTSO | IFCAP_VLAN_MTU;
686 
687 #define IGB_CAPS \
688 	IFCAP_TSO4 | IFCAP_TXCSUM | IFCAP_LRO | IFCAP_RXCSUM | IFCAP_VLAN_HWFILTER | IFCAP_WOL_MAGIC | \
689 	IFCAP_WOL_MCAST | IFCAP_WOL | IFCAP_VLAN_HWTSO | IFCAP_HWCSUM | IFCAP_VLAN_HWTAGGING | IFCAP_VLAN_HWCSUM | \
690 	IFCAP_VLAN_HWTSO | IFCAP_VLAN_MTU | IFCAP_TXCSUM_IPV6 | IFCAP_HWCSUM_IPV6 | IFCAP_JUMBO_MTU;
691 
692 /*********************************************************************
693  *  Device initialization routine
694  *
695  *  The attach entry point is called when the driver is being loaded.
696  *  This routine identifies the type of hardware, allocates all resources
697  *  and initializes the hardware.
698  *
699  *  return 0 on success, positive on failure
700  *********************************************************************/
701 
702 static int
703 em_if_attach_pre(if_ctx_t ctx)
704 {
705 	struct adapter *adapter;
706 	if_softc_ctx_t scctx;
707 	device_t dev;
708 	struct e1000_hw *hw;
709 	int error = 0;
710 
711 	INIT_DEBUGOUT("em_if_attach_pre begin");
712 	dev = iflib_get_dev(ctx);
713 	adapter = iflib_get_softc(ctx);
714 
715 	if (resource_disabled("em", device_get_unit(dev))) {
716 		device_printf(dev, "Disabled by device hint\n");
717 		return (ENXIO);
718 	}
719 
720 	adapter->ctx = ctx;
721 	adapter->dev = adapter->osdep.dev = dev;
722 	scctx = adapter->shared = iflib_get_softc_ctx(ctx);
723 	adapter->media = iflib_get_media(ctx);
724 	hw = &adapter->hw;
725 
726 	adapter->tx_process_limit = scctx->isc_ntxd[0];
727 
728 	/* SYSCTL stuff */
729 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
730 	    SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
731 	    OID_AUTO, "nvm", CTLTYPE_INT|CTLFLAG_RW, adapter, 0,
732 	    em_sysctl_nvm_info, "I", "NVM Information");
733 
734 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
735 	    SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
736 	    OID_AUTO, "debug", CTLTYPE_INT|CTLFLAG_RW, adapter, 0,
737 	    em_sysctl_debug_info, "I", "Debug Information");
738 
739 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
740 	    SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
741 	    OID_AUTO, "fc", CTLTYPE_INT|CTLFLAG_RW, adapter, 0,
742 	    em_set_flowcntl, "I", "Flow Control");
743 
744 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
745 	    SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
746 	    OID_AUTO, "reg_dump", CTLTYPE_STRING | CTLFLAG_RD, adapter, 0,
747 	    em_get_regs, "A", "Dump Registers");
748 
749 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
750 	    SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
751 	    OID_AUTO, "rs_dump", CTLTYPE_INT | CTLFLAG_RW, adapter, 0,
752 	    em_get_rs, "I", "Dump RS indexes");
753 
754 	/* Determine hardware and mac info */
755 	em_identify_hardware(ctx);
756 
757 	/* Set isc_msix_bar */
758 	scctx->isc_msix_bar = PCIR_BAR(EM_MSIX_BAR);
759 	scctx->isc_tx_nsegments = EM_MAX_SCATTER;
760 	scctx->isc_tx_tso_segments_max = scctx->isc_tx_nsegments;
761 	scctx->isc_tx_tso_size_max = EM_TSO_SIZE;
762 	scctx->isc_tx_tso_segsize_max = EM_TSO_SEG_SIZE;
763 	scctx->isc_nrxqsets_max = scctx->isc_ntxqsets_max = em_set_num_queues(ctx);
764 	device_printf(dev, "attach_pre capping queues at %d\n", scctx->isc_ntxqsets_max);
765 
766 	scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP | CSUM_IP_TSO;
767 
768 
769 	if (adapter->hw.mac.type >= igb_mac_min) {
770 		int try_second_bar;
771 
772 		scctx->isc_txqsizes[0] = roundup2(scctx->isc_ntxd[0] * sizeof(union e1000_adv_tx_desc), EM_DBA_ALIGN);
773 		scctx->isc_rxqsizes[0] = roundup2(scctx->isc_nrxd[0] * sizeof(union e1000_adv_rx_desc), EM_DBA_ALIGN);
774 		scctx->isc_txd_size[0] = sizeof(union e1000_adv_tx_desc);
775 		scctx->isc_rxd_size[0] = sizeof(union e1000_adv_rx_desc);
776 		scctx->isc_txrx = &igb_txrx;
777 		scctx->isc_capenable = IGB_CAPS;
778 		scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP | CSUM_TSO | CSUM_IP6_TCP \
779 			| CSUM_IP6_UDP | CSUM_IP6_TCP;
780 		if (adapter->hw.mac.type != e1000_82575)
781 			scctx->isc_tx_csum_flags |= CSUM_SCTP | CSUM_IP6_SCTP;
782 
783 		/*
784 		** Some new devices, as with ixgbe, now may
785 		** use a different BAR, so we need to keep
786 		** track of which is used.
787 		*/
788 		try_second_bar = pci_read_config(dev, scctx->isc_msix_bar, 4);
789 		if (try_second_bar == 0)
790 			scctx->isc_msix_bar += 4;
791 
792 	} else if (adapter->hw.mac.type >= em_mac_min) {
793 		scctx->isc_txqsizes[0] = roundup2(scctx->isc_ntxd[0]* sizeof(struct e1000_tx_desc), EM_DBA_ALIGN);
794 		scctx->isc_rxqsizes[0] = roundup2(scctx->isc_nrxd[0] * sizeof(union e1000_rx_desc_extended), EM_DBA_ALIGN);
795 		scctx->isc_txd_size[0] = sizeof(struct e1000_tx_desc);
796 		scctx->isc_rxd_size[0] = sizeof(union e1000_rx_desc_extended);
797 		scctx->isc_txrx = &em_txrx;
798 		scctx->isc_capenable = EM_CAPS;
799 		scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP | CSUM_IP_TSO;
800 	} else {
801 		scctx->isc_txqsizes[0] = roundup2((scctx->isc_ntxd[0] + 1) * sizeof(struct e1000_tx_desc), EM_DBA_ALIGN);
802 		scctx->isc_rxqsizes[0] = roundup2((scctx->isc_nrxd[0] + 1) * sizeof(struct e1000_rx_desc), EM_DBA_ALIGN);
803 		scctx->isc_txd_size[0] = sizeof(struct e1000_tx_desc);
804 		scctx->isc_rxd_size[0] = sizeof(struct e1000_rx_desc);
805 		scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP | CSUM_IP_TSO;
806 		scctx->isc_txrx = &lem_txrx;
807 		scctx->isc_capenable = EM_CAPS;
808 		if (adapter->hw.mac.type < e1000_82543)
809 			scctx->isc_capenable &= ~(IFCAP_HWCSUM|IFCAP_VLAN_HWCSUM);
810 		scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP | CSUM_IP_TSO;
811 		scctx->isc_msix_bar = 0;
812 	}
813 
814 	/* Setup PCI resources */
815 	if (em_allocate_pci_resources(ctx)) {
816 		device_printf(dev, "Allocation of PCI resources failed\n");
817 		error = ENXIO;
818 		goto err_pci;
819 	}
820 
821 	/*
822 	** For ICH8 and family we need to
823 	** map the flash memory, and this
824 	** must happen after the MAC is
825 	** identified
826 	*/
827 	if ((hw->mac.type == e1000_ich8lan) ||
828 	    (hw->mac.type == e1000_ich9lan) ||
829 	    (hw->mac.type == e1000_ich10lan) ||
830 	    (hw->mac.type == e1000_pchlan) ||
831 	    (hw->mac.type == e1000_pch2lan) ||
832 	    (hw->mac.type == e1000_pch_lpt)) {
833 		int rid = EM_BAR_TYPE_FLASH;
834 		adapter->flash = bus_alloc_resource_any(dev,
835 		    SYS_RES_MEMORY, &rid, RF_ACTIVE);
836 		if (adapter->flash == NULL) {
837 			device_printf(dev, "Mapping of Flash failed\n");
838 			error = ENXIO;
839 			goto err_pci;
840 		}
841 		/* This is used in the shared code */
842 		hw->flash_address = (u8 *)adapter->flash;
843 		adapter->osdep.flash_bus_space_tag =
844 		    rman_get_bustag(adapter->flash);
845 		adapter->osdep.flash_bus_space_handle =
846 		    rman_get_bushandle(adapter->flash);
847 	}
848 	/*
849 	** In the new SPT device flash is not  a
850 	** separate BAR, rather it is also in BAR0,
851 	** so use the same tag and an offset handle for the
852 	** FLASH read/write macros in the shared code.
853 	*/
854 	else if (hw->mac.type == e1000_pch_spt) {
855 		adapter->osdep.flash_bus_space_tag =
856 		    adapter->osdep.mem_bus_space_tag;
857 		adapter->osdep.flash_bus_space_handle =
858 		    adapter->osdep.mem_bus_space_handle
859 		    + E1000_FLASH_BASE_ADDR;
860 	}
861 
862 	/* Do Shared Code initialization */
863 	error = e1000_setup_init_funcs(hw, TRUE);
864 	if (error) {
865 		device_printf(dev, "Setup of Shared code failed, error %d\n",
866 		    error);
867 		error = ENXIO;
868 		goto err_pci;
869 	}
870 
871 	em_setup_msix(ctx);
872 	e1000_get_bus_info(hw);
873 
874 	/* Set up some sysctls for the tunable interrupt delays */
875 	em_add_int_delay_sysctl(adapter, "rx_int_delay",
876 	    "receive interrupt delay in usecs", &adapter->rx_int_delay,
877 	    E1000_REGISTER(hw, E1000_RDTR), em_rx_int_delay_dflt);
878 	em_add_int_delay_sysctl(adapter, "tx_int_delay",
879 	    "transmit interrupt delay in usecs", &adapter->tx_int_delay,
880 	    E1000_REGISTER(hw, E1000_TIDV), em_tx_int_delay_dflt);
881 	em_add_int_delay_sysctl(adapter, "rx_abs_int_delay",
882 	    "receive interrupt delay limit in usecs",
883 	    &adapter->rx_abs_int_delay,
884 	    E1000_REGISTER(hw, E1000_RADV),
885 	    em_rx_abs_int_delay_dflt);
886 	em_add_int_delay_sysctl(adapter, "tx_abs_int_delay",
887 	    "transmit interrupt delay limit in usecs",
888 	    &adapter->tx_abs_int_delay,
889 	    E1000_REGISTER(hw, E1000_TADV),
890 	    em_tx_abs_int_delay_dflt);
891 	em_add_int_delay_sysctl(adapter, "itr",
892 	    "interrupt delay limit in usecs/4",
893 	    &adapter->tx_itr,
894 	    E1000_REGISTER(hw, E1000_ITR),
895 	    DEFAULT_ITR);
896 
897 	hw->mac.autoneg = DO_AUTO_NEG;
898 	hw->phy.autoneg_wait_to_complete = FALSE;
899 	hw->phy.autoneg_advertised = AUTONEG_ADV_DEFAULT;
900 
901 	if (adapter->hw.mac.type < em_mac_min) {
902 		e1000_init_script_state_82541(&adapter->hw, TRUE);
903 		e1000_set_tbi_compatibility_82543(&adapter->hw, TRUE);
904 	}
905 	/* Copper options */
906 	if (hw->phy.media_type == e1000_media_type_copper) {
907 		hw->phy.mdix = AUTO_ALL_MODES;
908 		hw->phy.disable_polarity_correction = FALSE;
909 		hw->phy.ms_type = EM_MASTER_SLAVE;
910 	}
911 
912 	/*
913 	 * Set the frame limits assuming
914 	 * standard ethernet sized frames.
915 	 */
916 	scctx->isc_max_frame_size = adapter->hw.mac.max_frame_size =
917 	    ETHERMTU + ETHER_HDR_LEN + ETHERNET_FCS_SIZE;
918 
919 	/*
920 	 * This controls when hardware reports transmit completion
921 	 * status.
922 	 */
923 	hw->mac.report_tx_early = 1;
924 
925 	/* Allocate multicast array memory. */
926 	adapter->mta = malloc(sizeof(u8) * ETH_ADDR_LEN *
927 	    MAX_NUM_MULTICAST_ADDRESSES, M_DEVBUF, M_NOWAIT);
928 	if (adapter->mta == NULL) {
929 		device_printf(dev, "Can not allocate multicast setup array\n");
930 		error = ENOMEM;
931 		goto err_late;
932 	}
933 
934 	/* Check SOL/IDER usage */
935 	if (e1000_check_reset_block(hw))
936 		device_printf(dev, "PHY reset is blocked"
937 			      " due to SOL/IDER session.\n");
938 
939 	/* Sysctl for setting Energy Efficient Ethernet */
940 	hw->dev_spec.ich8lan.eee_disable = eee_setting;
941 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
942 	    SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
943 	    OID_AUTO, "eee_control", CTLTYPE_INT|CTLFLAG_RW,
944 	    adapter, 0, em_sysctl_eee, "I",
945 	    "Disable Energy Efficient Ethernet");
946 
947 	/*
948 	** Start from a known state, this is
949 	** important in reading the nvm and
950 	** mac from that.
951 	*/
952 	e1000_reset_hw(hw);
953 
954 	/* Make sure we have a good EEPROM before we read from it */
955 	if (e1000_validate_nvm_checksum(hw) < 0) {
956 		/*
957 		** Some PCI-E parts fail the first check due to
958 		** the link being in sleep state, call it again,
959 		** if it fails a second time its a real issue.
960 		*/
961 		if (e1000_validate_nvm_checksum(hw) < 0) {
962 			device_printf(dev,
963 			    "The EEPROM Checksum Is Not Valid\n");
964 			error = EIO;
965 			goto err_late;
966 		}
967 	}
968 
969 	/* Copy the permanent MAC address out of the EEPROM */
970 	if (e1000_read_mac_addr(hw) < 0) {
971 		device_printf(dev, "EEPROM read error while reading MAC"
972 			      " address\n");
973 		error = EIO;
974 		goto err_late;
975 	}
976 
977 	if (!em_is_valid_ether_addr(hw->mac.addr)) {
978 		device_printf(dev, "Invalid MAC address\n");
979 		error = EIO;
980 		goto err_late;
981 	}
982 
983 	/* Disable ULP support */
984 	e1000_disable_ulp_lpt_lp(hw, TRUE);
985 
986 	/*
987 	 * Get Wake-on-Lan and Management info for later use
988 	 */
989 	em_get_wakeup(ctx);
990 
991 	iflib_set_mac(ctx, hw->mac.addr);
992 
993 	return (0);
994 
995 err_late:
996 	em_release_hw_control(adapter);
997 err_pci:
998 	em_free_pci_resources(ctx);
999 	free(adapter->mta, M_DEVBUF);
1000 
1001 	return (error);
1002 }
1003 
1004 static int
1005 em_if_attach_post(if_ctx_t ctx)
1006 {
1007 	struct adapter *adapter = iflib_get_softc(ctx);
1008 	struct e1000_hw *hw = &adapter->hw;
1009 	int error = 0;
1010 
1011 	/* Setup OS specific network interface */
1012 	error = em_setup_interface(ctx);
1013 	if (error != 0) {
1014 		goto err_late;
1015 	}
1016 
1017 	em_reset(ctx);
1018 
1019 	/* Initialize statistics */
1020 	em_update_stats_counters(adapter);
1021 	hw->mac.get_link_status = 1;
1022 	em_if_update_admin_status(ctx);
1023 	em_add_hw_stats(adapter);
1024 
1025 	/* Non-AMT based hardware can now take control from firmware */
1026 	if (adapter->has_manage && !adapter->has_amt)
1027 		em_get_hw_control(adapter);
1028 
1029 	INIT_DEBUGOUT("em_if_attach_post: end");
1030 
1031 	return (error);
1032 
1033 err_late:
1034 	em_release_hw_control(adapter);
1035 	em_free_pci_resources(ctx);
1036 	em_if_queues_free(ctx);
1037 	free(adapter->mta, M_DEVBUF);
1038 
1039 	return (error);
1040 }
1041 
1042 /*********************************************************************
1043  *  Device removal routine
1044  *
1045  *  The detach entry point is called when the driver is being removed.
1046  *  This routine stops the adapter and deallocates all the resources
1047  *  that were allocated for driver operation.
1048  *
1049  *  return 0 on success, positive on failure
1050  *********************************************************************/
1051 
1052 static int
1053 em_if_detach(if_ctx_t ctx)
1054 {
1055 	struct adapter	*adapter = iflib_get_softc(ctx);
1056 
1057 	INIT_DEBUGOUT("em_detach: begin");
1058 
1059 	e1000_phy_hw_reset(&adapter->hw);
1060 
1061 	em_release_manageability(adapter);
1062 	em_release_hw_control(adapter);
1063 	em_free_pci_resources(ctx);
1064 
1065 	return (0);
1066 }
1067 
1068 /*********************************************************************
1069  *
1070  *  Shutdown entry point
1071  *
1072  **********************************************************************/
1073 
1074 static int
1075 em_if_shutdown(if_ctx_t ctx)
1076 {
1077 	return em_if_suspend(ctx);
1078 }
1079 
1080 /*
1081  * Suspend/resume device methods.
1082  */
1083 static int
1084 em_if_suspend(if_ctx_t ctx)
1085 {
1086 	struct adapter *adapter = iflib_get_softc(ctx);
1087 
1088 	em_release_manageability(adapter);
1089 	em_release_hw_control(adapter);
1090 	em_enable_wakeup(ctx);
1091 	return (0);
1092 }
1093 
1094 static int
1095 em_if_resume(if_ctx_t ctx)
1096 {
1097 	struct adapter *adapter = iflib_get_softc(ctx);
1098 
1099 	if (adapter->hw.mac.type == e1000_pch2lan)
1100 		e1000_resume_workarounds_pchlan(&adapter->hw);
1101 	em_if_init(ctx);
1102 	em_init_manageability(adapter);
1103 
1104 	return(0);
1105 }
1106 
1107 static int
1108 em_if_mtu_set(if_ctx_t ctx, uint32_t mtu)
1109 {
1110 	int max_frame_size;
1111 	struct adapter *adapter = iflib_get_softc(ctx);
1112 	if_softc_ctx_t scctx = iflib_get_softc_ctx(ctx);
1113 
1114 	 IOCTL_DEBUGOUT("ioctl rcv'd: SIOCSIFMTU (Set Interface MTU)");
1115 
1116 	switch (adapter->hw.mac.type) {
1117 	case e1000_82571:
1118 	case e1000_82572:
1119 	case e1000_ich9lan:
1120 	case e1000_ich10lan:
1121 	case e1000_pch2lan:
1122 	case e1000_pch_lpt:
1123 	case e1000_pch_spt:
1124 	case e1000_82574:
1125 	case e1000_82583:
1126 	case e1000_80003es2lan:
1127 		/* 9K Jumbo Frame size */
1128 		max_frame_size = 9234;
1129 		break;
1130 	case e1000_pchlan:
1131 		max_frame_size = 4096;
1132 		break;
1133 	case e1000_82542:
1134 	case e1000_ich8lan:
1135 		/* Adapters that do not support jumbo frames */
1136 		max_frame_size = ETHER_MAX_LEN;
1137 		break;
1138 	default:
1139 		if (adapter->hw.mac.type >= igb_mac_min)
1140 			max_frame_size = 9234;
1141 		else /* lem */
1142 			max_frame_size = MAX_JUMBO_FRAME_SIZE;
1143 	}
1144 	if (mtu > max_frame_size - ETHER_HDR_LEN - ETHER_CRC_LEN) {
1145 		return (EINVAL);
1146 	}
1147 
1148 	scctx->isc_max_frame_size = adapter->hw.mac.max_frame_size =
1149 	    mtu + ETHER_HDR_LEN + ETHER_CRC_LEN;
1150 	return (0);
1151 }
1152 
1153 /*********************************************************************
1154  *  Init entry point
1155  *
1156  *  This routine is used in two ways. It is used by the stack as
1157  *  init entry point in network interface structure. It is also used
1158  *  by the driver as a hw/sw initialization routine to get to a
1159  *  consistent state.
1160  *
1161  *  return 0 on success, positive on failure
1162  **********************************************************************/
1163 
1164 static void
1165 em_if_init(if_ctx_t ctx)
1166 {
1167 	struct adapter *adapter = iflib_get_softc(ctx);
1168 	struct ifnet *ifp = iflib_get_ifp(ctx);
1169 	struct em_tx_queue *tx_que;
1170 	int i;
1171 	INIT_DEBUGOUT("em_if_init: begin");
1172 
1173 	/* Get the latest mac address, User can use a LAA */
1174 	bcopy(if_getlladdr(ifp), adapter->hw.mac.addr,
1175 	    ETHER_ADDR_LEN);
1176 
1177 	/* Put the address into the Receive Address Array */
1178 	e1000_rar_set(&adapter->hw, adapter->hw.mac.addr, 0);
1179 
1180 	/*
1181 	 * With the 82571 adapter, RAR[0] may be overwritten
1182 	 * when the other port is reset, we make a duplicate
1183 	 * in RAR[14] for that eventuality, this assures
1184 	 * the interface continues to function.
1185 	 */
1186 	if (adapter->hw.mac.type == e1000_82571) {
1187 		e1000_set_laa_state_82571(&adapter->hw, TRUE);
1188 		e1000_rar_set(&adapter->hw, adapter->hw.mac.addr,
1189 		    E1000_RAR_ENTRIES - 1);
1190 	}
1191 
1192 
1193 	/* Initialize the hardware */
1194 	em_reset(ctx);
1195 	em_if_update_admin_status(ctx);
1196 
1197 	for (i = 0, tx_que = adapter->tx_queues; i < adapter->tx_num_queues; i++, tx_que++) {
1198 		struct tx_ring *txr = &tx_que->txr;
1199 
1200 		txr->tx_rs_cidx = txr->tx_rs_pidx = txr->tx_cidx_processed = 0;
1201 	}
1202 
1203 	/* Setup VLAN support, basic and offload if available */
1204 	E1000_WRITE_REG(&adapter->hw, E1000_VET, ETHERTYPE_VLAN);
1205 
1206 	/* Clear bad data from Rx FIFOs */
1207 	if (adapter->hw.mac.type >= igb_mac_min)
1208 		e1000_rx_fifo_flush_82575(&adapter->hw);
1209 
1210 	/* Configure for OS presence */
1211 	em_init_manageability(adapter);
1212 
1213 	/* Prepare transmit descriptors and buffers */
1214 	em_initialize_transmit_unit(ctx);
1215 
1216 	/* Setup Multicast table */
1217 	em_if_multi_set(ctx);
1218 
1219 	/*
1220 	 * Figure out the desired mbuf
1221 	 * pool for doing jumbos
1222 	 */
1223 	if (adapter->hw.mac.max_frame_size <= 2048)
1224 		adapter->rx_mbuf_sz = MCLBYTES;
1225 #ifndef CONTIGMALLOC_WORKS
1226 	else
1227 		adapter->rx_mbuf_sz = MJUMPAGESIZE;
1228 #else
1229 	else if (adapter->hw.mac.max_frame_size <= 4096)
1230 		adapter->rx_mbuf_sz = MJUMPAGESIZE;
1231 	else
1232 		adapter->rx_mbuf_sz = MJUM9BYTES;
1233 #endif
1234 	em_initialize_receive_unit(ctx);
1235 
1236 	/* Use real VLAN Filter support? */
1237 	if (if_getcapenable(ifp) & IFCAP_VLAN_HWTAGGING) {
1238 		if (if_getcapenable(ifp) & IFCAP_VLAN_HWFILTER)
1239 			/* Use real VLAN Filter support */
1240 			em_setup_vlan_hw_support(adapter);
1241 		else {
1242 			u32 ctrl;
1243 			ctrl = E1000_READ_REG(&adapter->hw, E1000_CTRL);
1244 			ctrl |= E1000_CTRL_VME;
1245 			E1000_WRITE_REG(&adapter->hw, E1000_CTRL, ctrl);
1246 		}
1247 	}
1248 
1249 	/* Don't lose promiscuous settings */
1250 	em_if_set_promisc(ctx, IFF_PROMISC);
1251 	e1000_clear_hw_cntrs_base_generic(&adapter->hw);
1252 
1253 	/* MSI/X configuration for 82574 */
1254 	if (adapter->hw.mac.type == e1000_82574) {
1255 		int tmp = E1000_READ_REG(&adapter->hw, E1000_CTRL_EXT);
1256 
1257 		tmp |= E1000_CTRL_EXT_PBA_CLR;
1258 		E1000_WRITE_REG(&adapter->hw, E1000_CTRL_EXT, tmp);
1259 		/* Set the IVAR - interrupt vector routing. */
1260 		E1000_WRITE_REG(&adapter->hw, E1000_IVAR, adapter->ivars);
1261 	} else if (adapter->intr_type == IFLIB_INTR_MSIX) /* Set up queue routing */
1262 		igb_configure_queues(adapter);
1263 
1264 	/* this clears any pending interrupts */
1265 	E1000_READ_REG(&adapter->hw, E1000_ICR);
1266 	E1000_WRITE_REG(&adapter->hw, E1000_ICS, E1000_ICS_LSC);
1267 
1268 	/* AMT based hardware can now take control from firmware */
1269 	if (adapter->has_manage && adapter->has_amt)
1270 		em_get_hw_control(adapter);
1271 
1272 	/* Set Energy Efficient Ethernet */
1273 	if (adapter->hw.mac.type >= igb_mac_min &&
1274 	    adapter->hw.phy.media_type == e1000_media_type_copper) {
1275 		if (adapter->hw.mac.type == e1000_i354)
1276 			e1000_set_eee_i354(&adapter->hw, TRUE, TRUE);
1277 		else
1278 			e1000_set_eee_i350(&adapter->hw, TRUE, TRUE);
1279 	}
1280 }
1281 
1282 /*********************************************************************
1283  *
1284  *  Fast Legacy/MSI Combined Interrupt Service routine
1285  *
1286  *********************************************************************/
1287 int
1288 em_intr(void *arg)
1289 {
1290 	struct adapter *adapter = arg;
1291 	if_ctx_t ctx = adapter->ctx;
1292 	u32 reg_icr;
1293 
1294 	reg_icr = E1000_READ_REG(&adapter->hw, E1000_ICR);
1295 
1296 	if (adapter->intr_type != IFLIB_INTR_LEGACY)
1297 		goto skip_stray;
1298 	/* Hot eject? */
1299 	if (reg_icr == 0xffffffff)
1300 		return FILTER_STRAY;
1301 
1302 	/* Definitely not our interrupt. */
1303 	if (reg_icr == 0x0)
1304 		return FILTER_STRAY;
1305 
1306 	/*
1307 	 * Starting with the 82571 chip, bit 31 should be used to
1308 	 * determine whether the interrupt belongs to us.
1309 	 */
1310 	if (adapter->hw.mac.type >= e1000_82571 &&
1311 	    (reg_icr & E1000_ICR_INT_ASSERTED) == 0)
1312 		return FILTER_STRAY;
1313 
1314 skip_stray:
1315 	/* Link status change */
1316 	if (reg_icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) {
1317 		adapter->hw.mac.get_link_status = 1;
1318 		iflib_admin_intr_deferred(ctx);
1319 	}
1320 
1321 	if (reg_icr & E1000_ICR_RXO)
1322 		adapter->rx_overruns++;
1323 
1324 	return (FILTER_SCHEDULE_THREAD);
1325 }
1326 
1327 static void
1328 igb_rx_enable_queue(struct adapter *adapter, struct em_rx_queue *rxq)
1329 {
1330 	E1000_WRITE_REG(&adapter->hw, E1000_EIMS, rxq->eims);
1331 }
1332 
1333 static void
1334 em_rx_enable_queue(struct adapter *adapter, struct em_rx_queue *rxq)
1335 {
1336 	E1000_WRITE_REG(&adapter->hw, E1000_IMS, rxq->eims);
1337 }
1338 
1339 static void
1340 igb_tx_enable_queue(struct adapter *adapter, struct em_tx_queue *txq)
1341 {
1342 	E1000_WRITE_REG(&adapter->hw, E1000_EIMS, txq->eims);
1343 }
1344 
1345 static void
1346 em_tx_enable_queue(struct adapter *adapter, struct em_tx_queue *txq)
1347 {
1348 	E1000_WRITE_REG(&adapter->hw, E1000_IMS, txq->eims);
1349 }
1350 
1351 static int
1352 em_if_rx_queue_intr_enable(if_ctx_t ctx, uint16_t rxqid)
1353 {
1354 	struct adapter *adapter = iflib_get_softc(ctx);
1355 	struct em_rx_queue *rxq = &adapter->rx_queues[rxqid];
1356 
1357 	if (adapter->hw.mac.type >= igb_mac_min)
1358 		igb_rx_enable_queue(adapter, rxq);
1359 	else
1360 		em_rx_enable_queue(adapter, rxq);
1361 	return (0);
1362 }
1363 
1364 static int
1365 em_if_tx_queue_intr_enable(if_ctx_t ctx, uint16_t txqid)
1366 {
1367 	struct adapter *adapter = iflib_get_softc(ctx);
1368 	struct em_tx_queue *txq = &adapter->tx_queues[txqid];
1369 
1370 	if (adapter->hw.mac.type >= igb_mac_min)
1371 		igb_tx_enable_queue(adapter, txq);
1372 	else
1373 		em_tx_enable_queue(adapter, txq);
1374 	return (0);
1375 }
1376 
1377 /*********************************************************************
1378  *
1379  *  MSIX RX Interrupt Service routine
1380  *
1381  **********************************************************************/
1382 static int
1383 em_msix_que(void *arg)
1384 {
1385 	struct em_rx_queue *que = arg;
1386 
1387 	++que->irqs;
1388 
1389 	return (FILTER_SCHEDULE_THREAD);
1390 }
1391 
1392 /*********************************************************************
1393  *
1394  *  MSIX Link Fast Interrupt Service routine
1395  *
1396  **********************************************************************/
1397 static int
1398 em_msix_link(void *arg)
1399 {
1400 	struct adapter *adapter = arg;
1401 	u32 reg_icr;
1402 
1403 	++adapter->link_irq;
1404 	MPASS(adapter->hw.back != NULL);
1405 	reg_icr = E1000_READ_REG(&adapter->hw, E1000_ICR);
1406 
1407 	if (reg_icr & E1000_ICR_RXO)
1408 		adapter->rx_overruns++;
1409 
1410 	if (reg_icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) {
1411 		em_handle_link(adapter->ctx);
1412 	} else {
1413 		E1000_WRITE_REG(&adapter->hw, E1000_IMS,
1414 				EM_MSIX_LINK | E1000_IMS_LSC);
1415 		if (adapter->hw.mac.type >= igb_mac_min)
1416 			E1000_WRITE_REG(&adapter->hw, E1000_EIMS, adapter->link_mask);
1417 	}
1418 
1419 	/*
1420 	 * Because we must read the ICR for this interrupt
1421 	 * it may clear other causes using autoclear, for
1422 	 * this reason we simply create a soft interrupt
1423 	 * for all these vectors.
1424 	 */
1425 	if (reg_icr && adapter->hw.mac.type < igb_mac_min) {
1426 		E1000_WRITE_REG(&adapter->hw,
1427 			E1000_ICS, adapter->ims);
1428 	}
1429 
1430 	return (FILTER_HANDLED);
1431 }
1432 
1433 static void
1434 em_handle_link(void *context)
1435 {
1436 	if_ctx_t ctx = context;
1437 	struct adapter *adapter = iflib_get_softc(ctx);
1438 
1439 	adapter->hw.mac.get_link_status = 1;
1440 	iflib_admin_intr_deferred(ctx);
1441 }
1442 
1443 
1444 /*********************************************************************
1445  *
1446  *  Media Ioctl callback
1447  *
1448  *  This routine is called whenever the user queries the status of
1449  *  the interface using ifconfig.
1450  *
1451  **********************************************************************/
1452 static void
1453 em_if_media_status(if_ctx_t ctx, struct ifmediareq *ifmr)
1454 {
1455 	struct adapter *adapter = iflib_get_softc(ctx);
1456 	u_char fiber_type = IFM_1000_SX;
1457 
1458 	INIT_DEBUGOUT("em_if_media_status: begin");
1459 
1460 	iflib_admin_intr_deferred(ctx);
1461 
1462 	ifmr->ifm_status = IFM_AVALID;
1463 	ifmr->ifm_active = IFM_ETHER;
1464 
1465 	if (!adapter->link_active) {
1466 		return;
1467 	}
1468 
1469 	ifmr->ifm_status |= IFM_ACTIVE;
1470 
1471 	if ((adapter->hw.phy.media_type == e1000_media_type_fiber) ||
1472 	    (adapter->hw.phy.media_type == e1000_media_type_internal_serdes)) {
1473 		if (adapter->hw.mac.type == e1000_82545)
1474 			fiber_type = IFM_1000_LX;
1475 		ifmr->ifm_active |= fiber_type | IFM_FDX;
1476 	} else {
1477 		switch (adapter->link_speed) {
1478 		case 10:
1479 			ifmr->ifm_active |= IFM_10_T;
1480 			break;
1481 		case 100:
1482 			ifmr->ifm_active |= IFM_100_TX;
1483 			break;
1484 		case 1000:
1485 			ifmr->ifm_active |= IFM_1000_T;
1486 			break;
1487 		}
1488 		if (adapter->link_duplex == FULL_DUPLEX)
1489 			ifmr->ifm_active |= IFM_FDX;
1490 		else
1491 			ifmr->ifm_active |= IFM_HDX;
1492 	}
1493 }
1494 
1495 /*********************************************************************
1496  *
1497  *  Media Ioctl callback
1498  *
1499  *  This routine is called when the user changes speed/duplex using
1500  *  media/mediopt option with ifconfig.
1501  *
1502  **********************************************************************/
1503 static int
1504 em_if_media_change(if_ctx_t ctx)
1505 {
1506 	struct adapter *adapter = iflib_get_softc(ctx);
1507 	struct ifmedia *ifm = iflib_get_media(ctx);
1508 
1509 	INIT_DEBUGOUT("em_if_media_change: begin");
1510 
1511 	if (IFM_TYPE(ifm->ifm_media) != IFM_ETHER)
1512 		return (EINVAL);
1513 
1514 	switch (IFM_SUBTYPE(ifm->ifm_media)) {
1515 	case IFM_AUTO:
1516 		adapter->hw.mac.autoneg = DO_AUTO_NEG;
1517 		adapter->hw.phy.autoneg_advertised = AUTONEG_ADV_DEFAULT;
1518 		break;
1519 	case IFM_1000_LX:
1520 	case IFM_1000_SX:
1521 	case IFM_1000_T:
1522 		adapter->hw.mac.autoneg = DO_AUTO_NEG;
1523 		adapter->hw.phy.autoneg_advertised = ADVERTISE_1000_FULL;
1524 		break;
1525 	case IFM_100_TX:
1526 		adapter->hw.mac.autoneg = FALSE;
1527 		adapter->hw.phy.autoneg_advertised = 0;
1528 		if ((ifm->ifm_media & IFM_GMASK) == IFM_FDX)
1529 			adapter->hw.mac.forced_speed_duplex = ADVERTISE_100_FULL;
1530 		else
1531 			adapter->hw.mac.forced_speed_duplex = ADVERTISE_100_HALF;
1532 		break;
1533 	case IFM_10_T:
1534 		adapter->hw.mac.autoneg = FALSE;
1535 		adapter->hw.phy.autoneg_advertised = 0;
1536 		if ((ifm->ifm_media & IFM_GMASK) == IFM_FDX)
1537 			adapter->hw.mac.forced_speed_duplex = ADVERTISE_10_FULL;
1538 		else
1539 			adapter->hw.mac.forced_speed_duplex = ADVERTISE_10_HALF;
1540 		break;
1541 	default:
1542 		device_printf(adapter->dev, "Unsupported media type\n");
1543 	}
1544 
1545 	em_if_init(ctx);
1546 
1547 	return (0);
1548 }
1549 
1550 static int
1551 em_if_set_promisc(if_ctx_t ctx, int flags)
1552 {
1553 	struct adapter *adapter = iflib_get_softc(ctx);
1554 	u32 reg_rctl;
1555 
1556 	em_disable_promisc(ctx);
1557 
1558 	reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
1559 
1560 	if (flags & IFF_PROMISC) {
1561 		reg_rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
1562 		/* Turn this on if you want to see bad packets */
1563 		if (em_debug_sbp)
1564 			reg_rctl |= E1000_RCTL_SBP;
1565 		E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1566 	} else if (flags & IFF_ALLMULTI) {
1567 		reg_rctl |= E1000_RCTL_MPE;
1568 		reg_rctl &= ~E1000_RCTL_UPE;
1569 		E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1570 	}
1571 	return (0);
1572 }
1573 
1574 static void
1575 em_disable_promisc(if_ctx_t ctx)
1576 {
1577 	struct adapter *adapter = iflib_get_softc(ctx);
1578 	struct ifnet *ifp = iflib_get_ifp(ctx);
1579 	u32 reg_rctl;
1580 	int mcnt = 0;
1581 
1582 	reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
1583 	reg_rctl &= (~E1000_RCTL_UPE);
1584 	if (if_getflags(ifp) & IFF_ALLMULTI)
1585 		mcnt = MAX_NUM_MULTICAST_ADDRESSES;
1586 	else
1587 		mcnt = if_multiaddr_count(ifp, MAX_NUM_MULTICAST_ADDRESSES);
1588 	/* Don't disable if in MAX groups */
1589 	if (mcnt < MAX_NUM_MULTICAST_ADDRESSES)
1590 		reg_rctl &=  (~E1000_RCTL_MPE);
1591 	reg_rctl &=  (~E1000_RCTL_SBP);
1592 	E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1593 }
1594 
1595 
1596 /*********************************************************************
1597  *  Multicast Update
1598  *
1599  *  This routine is called whenever multicast address list is updated.
1600  *
1601  **********************************************************************/
1602 
1603 static void
1604 em_if_multi_set(if_ctx_t ctx)
1605 {
1606 	struct adapter *adapter = iflib_get_softc(ctx);
1607 	struct ifnet *ifp = iflib_get_ifp(ctx);
1608 	u32 reg_rctl = 0;
1609 	u8  *mta; /* Multicast array memory */
1610 	int mcnt = 0;
1611 
1612 	IOCTL_DEBUGOUT("em_set_multi: begin");
1613 
1614 	mta = adapter->mta;
1615 	bzero(mta, sizeof(u8) * ETH_ADDR_LEN * MAX_NUM_MULTICAST_ADDRESSES);
1616 
1617 	if (adapter->hw.mac.type == e1000_82542 &&
1618 	    adapter->hw.revision_id == E1000_REVISION_2) {
1619 		reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
1620 		if (adapter->hw.bus.pci_cmd_word & CMD_MEM_WRT_INVALIDATE)
1621 			e1000_pci_clear_mwi(&adapter->hw);
1622 		reg_rctl |= E1000_RCTL_RST;
1623 		E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1624 		msec_delay(5);
1625 	}
1626 
1627 	if_multiaddr_array(ifp, mta, &mcnt, MAX_NUM_MULTICAST_ADDRESSES);
1628 
1629 	if (mcnt >= MAX_NUM_MULTICAST_ADDRESSES) {
1630 		reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
1631 		reg_rctl |= E1000_RCTL_MPE;
1632 		E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1633 	} else
1634 		e1000_update_mc_addr_list(&adapter->hw, mta, mcnt);
1635 
1636 	if (adapter->hw.mac.type == e1000_82542 &&
1637 	    adapter->hw.revision_id == E1000_REVISION_2) {
1638 		reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
1639 		reg_rctl &= ~E1000_RCTL_RST;
1640 		E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1641 		msec_delay(5);
1642 		if (adapter->hw.bus.pci_cmd_word & CMD_MEM_WRT_INVALIDATE)
1643 			e1000_pci_set_mwi(&adapter->hw);
1644 	}
1645 }
1646 
1647 
1648 /*********************************************************************
1649  *  Timer routine
1650  *
1651  *  This routine checks for link status and updates statistics.
1652  *
1653  **********************************************************************/
1654 
1655 static void
1656 em_if_timer(if_ctx_t ctx, uint16_t qid)
1657 {
1658 	struct adapter *adapter = iflib_get_softc(ctx);
1659 	struct em_rx_queue *que;
1660 	int i;
1661 	int trigger = 0;
1662 
1663 	if (qid != 0)
1664 		return;
1665 
1666 	iflib_admin_intr_deferred(ctx);
1667 	/* Reset LAA into RAR[0] on 82571 */
1668 	if ((adapter->hw.mac.type == e1000_82571) &&
1669 	    e1000_get_laa_state_82571(&adapter->hw))
1670 		e1000_rar_set(&adapter->hw, adapter->hw.mac.addr, 0);
1671 
1672 	if (adapter->hw.mac.type < em_mac_min)
1673 		lem_smartspeed(adapter);
1674 
1675 	/* Mask to use in the irq trigger */
1676 	if (adapter->intr_type == IFLIB_INTR_MSIX) {
1677 		for (i = 0, que = adapter->rx_queues; i < adapter->rx_num_queues; i++, que++)
1678 			trigger |= que->eims;
1679 	} else {
1680 		trigger = E1000_ICS_RXDMT0;
1681 	}
1682 }
1683 
1684 
1685 static void
1686 em_if_update_admin_status(if_ctx_t ctx)
1687 {
1688 	struct adapter *adapter = iflib_get_softc(ctx);
1689 	struct e1000_hw *hw = &adapter->hw;
1690 	struct ifnet *ifp = iflib_get_ifp(ctx);
1691 	device_t dev = iflib_get_dev(ctx);
1692 	u32 link_check, thstat, ctrl;
1693 
1694 	link_check = thstat = ctrl = 0;
1695 	/* Get the cached link value or read phy for real */
1696 	switch (hw->phy.media_type) {
1697 	case e1000_media_type_copper:
1698 		if (hw->mac.get_link_status) {
1699 			if (hw->mac.type == e1000_pch_spt)
1700 				msec_delay(50);
1701 			/* Do the work to read phy */
1702 			e1000_check_for_link(hw);
1703 			link_check = !hw->mac.get_link_status;
1704 			if (link_check) /* ESB2 fix */
1705 				e1000_cfg_on_link_up(hw);
1706 		} else {
1707 			link_check = TRUE;
1708 		}
1709 		break;
1710 	case e1000_media_type_fiber:
1711 		e1000_check_for_link(hw);
1712 		link_check = (E1000_READ_REG(hw, E1000_STATUS) &
1713 			    E1000_STATUS_LU);
1714 		break;
1715 	case e1000_media_type_internal_serdes:
1716 		e1000_check_for_link(hw);
1717 		link_check = adapter->hw.mac.serdes_has_link;
1718 		break;
1719 	/* VF device is type_unknown */
1720 	case e1000_media_type_unknown:
1721 		e1000_check_for_link(hw);
1722 		link_check = !hw->mac.get_link_status;
1723 		/* FALLTHROUGH */
1724 	default:
1725 		break;
1726 	}
1727 
1728 	/* Check for thermal downshift or shutdown */
1729 	if (hw->mac.type == e1000_i350) {
1730 		thstat = E1000_READ_REG(hw, E1000_THSTAT);
1731 		ctrl = E1000_READ_REG(hw, E1000_CTRL_EXT);
1732 	}
1733 
1734 	/* Now check for a transition */
1735 	if (link_check && (adapter->link_active == 0)) {
1736 		e1000_get_speed_and_duplex(hw, &adapter->link_speed,
1737 		    &adapter->link_duplex);
1738 		/* Check if we must disable SPEED_MODE bit on PCI-E */
1739 		if ((adapter->link_speed != SPEED_1000) &&
1740 		    ((hw->mac.type == e1000_82571) ||
1741 		    (hw->mac.type == e1000_82572))) {
1742 			int tarc0;
1743 			tarc0 = E1000_READ_REG(hw, E1000_TARC(0));
1744 			tarc0 &= ~TARC_SPEED_MODE_BIT;
1745 			E1000_WRITE_REG(hw, E1000_TARC(0), tarc0);
1746 		}
1747 		if (bootverbose)
1748 			device_printf(dev, "Link is up %d Mbps %s\n",
1749 			    adapter->link_speed,
1750 			    ((adapter->link_duplex == FULL_DUPLEX) ?
1751 			    "Full Duplex" : "Half Duplex"));
1752 		adapter->link_active = 1;
1753 		adapter->smartspeed = 0;
1754 		if_setbaudrate(ifp, adapter->link_speed * 1000000);
1755 		if ((ctrl & E1000_CTRL_EXT_LINK_MODE_GMII) &&
1756 		    (thstat & E1000_THSTAT_LINK_THROTTLE))
1757 			device_printf(dev, "Link: thermal downshift\n");
1758 		/* Delay Link Up for Phy update */
1759 		if (((hw->mac.type == e1000_i210) ||
1760 		    (hw->mac.type == e1000_i211)) &&
1761 		    (hw->phy.id == I210_I_PHY_ID))
1762 			msec_delay(I210_LINK_DELAY);
1763 		/* Reset if the media type changed. */
1764 		if ((hw->dev_spec._82575.media_changed) &&
1765 			(adapter->hw.mac.type >= igb_mac_min)) {
1766 			hw->dev_spec._82575.media_changed = false;
1767 			adapter->flags |= IGB_MEDIA_RESET;
1768 			em_reset(ctx);
1769 		}
1770 		iflib_link_state_change(ctx, LINK_STATE_UP, ifp->if_baudrate);
1771 		printf("Link state changed to up\n");
1772 	} else if (!link_check && (adapter->link_active == 1)) {
1773 		if_setbaudrate(ifp, 0);
1774 		adapter->link_speed = 0;
1775 		adapter->link_duplex = 0;
1776 		if (bootverbose)
1777 			device_printf(dev, "Link is Down\n");
1778 		adapter->link_active = 0;
1779 		iflib_link_state_change(ctx, LINK_STATE_DOWN, ifp->if_baudrate);
1780 		printf("link state changed to down\n");
1781 	}
1782 	em_update_stats_counters(adapter);
1783 
1784 	E1000_WRITE_REG(&adapter->hw, E1000_IMS, EM_MSIX_LINK | E1000_IMS_LSC);
1785 }
1786 
1787 /*********************************************************************
1788  *
1789  *  This routine disables all traffic on the adapter by issuing a
1790  *  global reset on the MAC and deallocates TX/RX buffers.
1791  *
1792  *  This routine should always be called with BOTH the CORE
1793  *  and TX locks.
1794  **********************************************************************/
1795 
1796 static void
1797 em_if_stop(if_ctx_t ctx)
1798 {
1799 	struct adapter *adapter = iflib_get_softc(ctx);
1800 
1801 	INIT_DEBUGOUT("em_stop: begin");
1802 
1803 	e1000_reset_hw(&adapter->hw);
1804 	if (adapter->hw.mac.type >= e1000_82544)
1805 		E1000_WRITE_REG(&adapter->hw, E1000_WUFC, 0);
1806 
1807 	e1000_led_off(&adapter->hw);
1808 	e1000_cleanup_led(&adapter->hw);
1809 }
1810 
1811 
1812 /*********************************************************************
1813  *
1814  *  Determine hardware revision.
1815  *
1816  **********************************************************************/
1817 static void
1818 em_identify_hardware(if_ctx_t ctx)
1819 {
1820 	device_t dev = iflib_get_dev(ctx);
1821 	struct adapter *adapter = iflib_get_softc(ctx);
1822 
1823 	/* Make sure our PCI config space has the necessary stuff set */
1824 	adapter->hw.bus.pci_cmd_word = pci_read_config(dev, PCIR_COMMAND, 2);
1825 
1826 	/* Save off the information about this board */
1827 	adapter->hw.vendor_id = pci_get_vendor(dev);
1828 	adapter->hw.device_id = pci_get_device(dev);
1829 	adapter->hw.revision_id = pci_read_config(dev, PCIR_REVID, 1);
1830 	adapter->hw.subsystem_vendor_id =
1831 	    pci_read_config(dev, PCIR_SUBVEND_0, 2);
1832 	adapter->hw.subsystem_device_id =
1833 	    pci_read_config(dev, PCIR_SUBDEV_0, 2);
1834 
1835 	/* Do Shared Code Init and Setup */
1836 	if (e1000_set_mac_type(&adapter->hw)) {
1837 		device_printf(dev, "Setup init failure\n");
1838 		return;
1839 	}
1840 }
1841 
1842 static int
1843 em_allocate_pci_resources(if_ctx_t ctx)
1844 {
1845 	struct adapter *adapter = iflib_get_softc(ctx);
1846 	device_t dev = iflib_get_dev(ctx);
1847 	int rid, val;
1848 
1849 	rid = PCIR_BAR(0);
1850 	adapter->memory = bus_alloc_resource_any(dev, SYS_RES_MEMORY,
1851 	    &rid, RF_ACTIVE);
1852 	if (adapter->memory == NULL) {
1853 		device_printf(dev, "Unable to allocate bus resource: memory\n");
1854 		return (ENXIO);
1855 	}
1856 	adapter->osdep.mem_bus_space_tag = rman_get_bustag(adapter->memory);
1857 	adapter->osdep.mem_bus_space_handle =
1858 	    rman_get_bushandle(adapter->memory);
1859 	adapter->hw.hw_addr = (u8 *)&adapter->osdep.mem_bus_space_handle;
1860 
1861 	/* Only older adapters use IO mapping */
1862 	if (adapter->hw.mac.type < em_mac_min &&
1863 	    adapter->hw.mac.type > e1000_82543) {
1864 		/* Figure our where our IO BAR is ? */
1865 		for (rid = PCIR_BAR(0); rid < PCIR_CIS;) {
1866 			val = pci_read_config(dev, rid, 4);
1867 			if (EM_BAR_TYPE(val) == EM_BAR_TYPE_IO) {
1868 				adapter->io_rid = rid;
1869 				break;
1870 			}
1871 			rid += 4;
1872 			/* check for 64bit BAR */
1873 			if (EM_BAR_MEM_TYPE(val) == EM_BAR_MEM_TYPE_64BIT)
1874 				rid += 4;
1875 		}
1876 		if (rid >= PCIR_CIS) {
1877 			device_printf(dev, "Unable to locate IO BAR\n");
1878 			return (ENXIO);
1879 		}
1880 		adapter->ioport = bus_alloc_resource_any(dev,
1881 		    SYS_RES_IOPORT, &adapter->io_rid, RF_ACTIVE);
1882 		if (adapter->ioport == NULL) {
1883 			device_printf(dev, "Unable to allocate bus resource: "
1884 			    "ioport\n");
1885 			return (ENXIO);
1886 		}
1887 		adapter->hw.io_base = 0;
1888 		adapter->osdep.io_bus_space_tag =
1889 		    rman_get_bustag(adapter->ioport);
1890 		adapter->osdep.io_bus_space_handle =
1891 		    rman_get_bushandle(adapter->ioport);
1892 	}
1893 
1894 	adapter->hw.back = &adapter->osdep;
1895 
1896 	return (0);
1897 }
1898 
1899 /*********************************************************************
1900  *
1901  *  Setup the MSIX Interrupt handlers
1902  *
1903  **********************************************************************/
1904 static int
1905 em_if_msix_intr_assign(if_ctx_t ctx, int msix)
1906 {
1907 	struct adapter *adapter = iflib_get_softc(ctx);
1908 	struct em_rx_queue *rx_que = adapter->rx_queues;
1909 	struct em_tx_queue *tx_que = adapter->tx_queues;
1910 	int error, rid, i, vector = 0, rx_vectors;
1911 	char buf[16];
1912 
1913 	/* First set up ring resources */
1914 	for (i = 0; i < adapter->rx_num_queues; i++, rx_que++, vector++) {
1915 		rid = vector + 1;
1916 		snprintf(buf, sizeof(buf), "rxq%d", i);
1917 		error = iflib_irq_alloc_generic(ctx, &rx_que->que_irq, rid, IFLIB_INTR_RXTX, em_msix_que, rx_que, rx_que->me, buf);
1918 		if (error) {
1919 			device_printf(iflib_get_dev(ctx), "Failed to allocate que int %d err: %d", i, error);
1920 			adapter->rx_num_queues = i + 1;
1921 			goto fail;
1922 		}
1923 
1924 		rx_que->msix =  vector;
1925 
1926 		/*
1927 		 * Set the bit to enable interrupt
1928 		 * in E1000_IMS -- bits 20 and 21
1929 		 * are for RX0 and RX1, note this has
1930 		 * NOTHING to do with the MSIX vector
1931 		 */
1932 		if (adapter->hw.mac.type == e1000_82574) {
1933 			rx_que->eims = 1 << (20 + i);
1934 			adapter->ims |= rx_que->eims;
1935 			adapter->ivars |= (8 | rx_que->msix) << (i * 4);
1936 		} else if (adapter->hw.mac.type == e1000_82575)
1937 			rx_que->eims = E1000_EICR_TX_QUEUE0 << vector;
1938 		else
1939 			rx_que->eims = 1 << vector;
1940 	}
1941 	rx_vectors = vector;
1942 
1943 	vector = 0;
1944 	for (i = 0; i < adapter->tx_num_queues; i++, tx_que++, vector++) {
1945 		rid = vector + 1;
1946 		snprintf(buf, sizeof(buf), "txq%d", i);
1947 		tx_que = &adapter->tx_queues[i];
1948 		iflib_softirq_alloc_generic(ctx, rid, IFLIB_INTR_TX, tx_que, tx_que->me, buf);
1949 
1950 		tx_que->msix = (vector % adapter->tx_num_queues);
1951 
1952 		/*
1953 		 * Set the bit to enable interrupt
1954 		 * in E1000_IMS -- bits 22 and 23
1955 		 * are for TX0 and TX1, note this has
1956 		 * NOTHING to do with the MSIX vector
1957 		 */
1958 		if (adapter->hw.mac.type == e1000_82574) {
1959 			tx_que->eims = 1 << (22 + i);
1960 			adapter->ims |= tx_que->eims;
1961 			adapter->ivars |= (8 | tx_que->msix) << (8 + (i * 4));
1962 		} else if (adapter->hw.mac.type == e1000_82575) {
1963 			tx_que->eims = E1000_EICR_TX_QUEUE0 << (i %  adapter->tx_num_queues);
1964 		} else {
1965 			tx_que->eims = 1 << (i %  adapter->tx_num_queues);
1966 		}
1967 	}
1968 
1969 	/* Link interrupt */
1970 	rid = rx_vectors + 1;
1971 	error = iflib_irq_alloc_generic(ctx, &adapter->irq, rid, IFLIB_INTR_ADMIN, em_msix_link, adapter, 0, "aq");
1972 
1973 	if (error) {
1974 		device_printf(iflib_get_dev(ctx), "Failed to register admin handler");
1975 		goto fail;
1976 	}
1977 	adapter->linkvec = rx_vectors;
1978 	if (adapter->hw.mac.type < igb_mac_min) {
1979 		adapter->ivars |=  (8 | rx_vectors) << 16;
1980 		adapter->ivars |= 0x80000000;
1981 	}
1982 	return (0);
1983 fail:
1984 	iflib_irq_free(ctx, &adapter->irq);
1985 	rx_que = adapter->rx_queues;
1986 	for (int i = 0; i < adapter->rx_num_queues; i++, rx_que++)
1987 		iflib_irq_free(ctx, &rx_que->que_irq);
1988 	return (error);
1989 }
1990 
1991 static void
1992 igb_configure_queues(struct adapter *adapter)
1993 {
1994 	struct e1000_hw *hw = &adapter->hw;
1995 	struct em_rx_queue *rx_que;
1996 	struct em_tx_queue *tx_que;
1997 	u32 tmp, ivar = 0, newitr = 0;
1998 
1999 	/* First turn on RSS capability */
2000 	if (adapter->hw.mac.type != e1000_82575)
2001 		E1000_WRITE_REG(hw, E1000_GPIE,
2002 		    E1000_GPIE_MSIX_MODE | E1000_GPIE_EIAME |
2003 		    E1000_GPIE_PBA | E1000_GPIE_NSICR);
2004 
2005 	/* Turn on MSIX */
2006 	switch (adapter->hw.mac.type) {
2007 	case e1000_82580:
2008 	case e1000_i350:
2009 	case e1000_i354:
2010 	case e1000_i210:
2011 	case e1000_i211:
2012 	case e1000_vfadapt:
2013 	case e1000_vfadapt_i350:
2014 		/* RX entries */
2015 		for (int i = 0; i < adapter->rx_num_queues; i++) {
2016 			u32 index = i >> 1;
2017 			ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2018 			rx_que = &adapter->rx_queues[i];
2019 			if (i & 1) {
2020 				ivar &= 0xFF00FFFF;
2021 				ivar |= (rx_que->msix | E1000_IVAR_VALID) << 16;
2022 			} else {
2023 				ivar &= 0xFFFFFF00;
2024 				ivar |= rx_que->msix | E1000_IVAR_VALID;
2025 			}
2026 			E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2027 		}
2028 		/* TX entries */
2029 		for (int i = 0; i < adapter->tx_num_queues; i++) {
2030 			u32 index = i >> 1;
2031 			ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2032 			tx_que = &adapter->tx_queues[i];
2033 			if (i & 1) {
2034 				ivar &= 0x00FFFFFF;
2035 				ivar |= (tx_que->msix | E1000_IVAR_VALID) << 24;
2036 			} else {
2037 				ivar &= 0xFFFF00FF;
2038 				ivar |= (tx_que->msix | E1000_IVAR_VALID) << 8;
2039 			}
2040 			E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2041 			adapter->que_mask |= tx_que->eims;
2042 		}
2043 
2044 		/* And for the link interrupt */
2045 		ivar = (adapter->linkvec | E1000_IVAR_VALID) << 8;
2046 		adapter->link_mask = 1 << adapter->linkvec;
2047 		E1000_WRITE_REG(hw, E1000_IVAR_MISC, ivar);
2048 		break;
2049 	case e1000_82576:
2050 		/* RX entries */
2051 		for (int i = 0; i < adapter->rx_num_queues; i++) {
2052 			u32 index = i & 0x7; /* Each IVAR has two entries */
2053 			ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2054 			rx_que = &adapter->rx_queues[i];
2055 			if (i < 8) {
2056 				ivar &= 0xFFFFFF00;
2057 				ivar |= rx_que->msix | E1000_IVAR_VALID;
2058 			} else {
2059 				ivar &= 0xFF00FFFF;
2060 				ivar |= (rx_que->msix | E1000_IVAR_VALID) << 16;
2061 			}
2062 			E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2063 			adapter->que_mask |= rx_que->eims;
2064 		}
2065 		/* TX entries */
2066 		for (int i = 0; i < adapter->tx_num_queues; i++) {
2067 			u32 index = i & 0x7; /* Each IVAR has two entries */
2068 			ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2069 			tx_que = &adapter->tx_queues[i];
2070 			if (i < 8) {
2071 				ivar &= 0xFFFF00FF;
2072 				ivar |= (tx_que->msix | E1000_IVAR_VALID) << 8;
2073 			} else {
2074 				ivar &= 0x00FFFFFF;
2075 				ivar |= (tx_que->msix | E1000_IVAR_VALID) << 24;
2076 			}
2077 			E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2078 			adapter->que_mask |= tx_que->eims;
2079 		}
2080 
2081 		/* And for the link interrupt */
2082 		ivar = (adapter->linkvec | E1000_IVAR_VALID) << 8;
2083 		adapter->link_mask = 1 << adapter->linkvec;
2084 		E1000_WRITE_REG(hw, E1000_IVAR_MISC, ivar);
2085 		break;
2086 
2087 	case e1000_82575:
2088 		/* enable MSI-X support*/
2089 		tmp = E1000_READ_REG(hw, E1000_CTRL_EXT);
2090 		tmp |= E1000_CTRL_EXT_PBA_CLR;
2091 		/* Auto-Mask interrupts upon ICR read. */
2092 		tmp |= E1000_CTRL_EXT_EIAME;
2093 		tmp |= E1000_CTRL_EXT_IRCA;
2094 		E1000_WRITE_REG(hw, E1000_CTRL_EXT, tmp);
2095 
2096 		/* Queues */
2097 		for (int i = 0; i < adapter->rx_num_queues; i++) {
2098 			rx_que = &adapter->rx_queues[i];
2099 			tmp = E1000_EICR_RX_QUEUE0 << i;
2100 			tmp |= E1000_EICR_TX_QUEUE0 << i;
2101 			rx_que->eims = tmp;
2102 			E1000_WRITE_REG_ARRAY(hw, E1000_MSIXBM(0),
2103 			    i, rx_que->eims);
2104 			adapter->que_mask |= rx_que->eims;
2105 		}
2106 
2107 		/* Link */
2108 		E1000_WRITE_REG(hw, E1000_MSIXBM(adapter->linkvec),
2109 		    E1000_EIMS_OTHER);
2110 		adapter->link_mask |= E1000_EIMS_OTHER;
2111 	default:
2112 		break;
2113 	}
2114 
2115 	/* Set the starting interrupt rate */
2116 	if (em_max_interrupt_rate > 0)
2117 		newitr = (4000000 / em_max_interrupt_rate) & 0x7FFC;
2118 
2119 	if (hw->mac.type == e1000_82575)
2120 		newitr |= newitr << 16;
2121 	else
2122 		newitr |= E1000_EITR_CNT_IGNR;
2123 
2124 	for (int i = 0; i < adapter->rx_num_queues; i++) {
2125 		rx_que = &adapter->rx_queues[i];
2126 		E1000_WRITE_REG(hw, E1000_EITR(rx_que->msix), newitr);
2127 	}
2128 
2129 	return;
2130 }
2131 
2132 static void
2133 em_free_pci_resources(if_ctx_t ctx)
2134 {
2135 	struct adapter *adapter = iflib_get_softc(ctx);
2136 	struct em_rx_queue *que = adapter->rx_queues;
2137 	device_t dev = iflib_get_dev(ctx);
2138 
2139 	/* Release all msix queue resources */
2140 	if (adapter->intr_type == IFLIB_INTR_MSIX)
2141 		iflib_irq_free(ctx, &adapter->irq);
2142 
2143 	for (int i = 0; i < adapter->rx_num_queues; i++, que++) {
2144 		iflib_irq_free(ctx, &que->que_irq);
2145 	}
2146 
2147 	/* First release all the interrupt resources */
2148 	if (adapter->memory != NULL) {
2149 		bus_release_resource(dev, SYS_RES_MEMORY,
2150 				     PCIR_BAR(0), adapter->memory);
2151 		adapter->memory = NULL;
2152 	}
2153 
2154 	if (adapter->flash != NULL) {
2155 		bus_release_resource(dev, SYS_RES_MEMORY,
2156 				     EM_FLASH, adapter->flash);
2157 		adapter->flash = NULL;
2158 	}
2159 	if (adapter->ioport != NULL)
2160 		bus_release_resource(dev, SYS_RES_IOPORT,
2161 		    adapter->io_rid, adapter->ioport);
2162 }
2163 
2164 /* Setup MSI or MSI/X */
2165 static int
2166 em_setup_msix(if_ctx_t ctx)
2167 {
2168 	struct adapter *adapter = iflib_get_softc(ctx);
2169 
2170 	if (adapter->hw.mac.type == e1000_82574) {
2171 		em_enable_vectors_82574(ctx);
2172 	}
2173 	return (0);
2174 }
2175 
2176 /*********************************************************************
2177  *
2178  *  Initialize the hardware to a configuration
2179  *  as specified by the adapter structure.
2180  *
2181  **********************************************************************/
2182 
2183 static void
2184 lem_smartspeed(struct adapter *adapter)
2185 {
2186 	u16 phy_tmp;
2187 
2188 	if (adapter->link_active || (adapter->hw.phy.type != e1000_phy_igp) ||
2189 	    adapter->hw.mac.autoneg == 0 ||
2190 	    (adapter->hw.phy.autoneg_advertised & ADVERTISE_1000_FULL) == 0)
2191 		return;
2192 
2193 	if (adapter->smartspeed == 0) {
2194 		/* If Master/Slave config fault is asserted twice,
2195 		 * we assume back-to-back */
2196 		e1000_read_phy_reg(&adapter->hw, PHY_1000T_STATUS, &phy_tmp);
2197 		if (!(phy_tmp & SR_1000T_MS_CONFIG_FAULT))
2198 			return;
2199 		e1000_read_phy_reg(&adapter->hw, PHY_1000T_STATUS, &phy_tmp);
2200 		if (phy_tmp & SR_1000T_MS_CONFIG_FAULT) {
2201 			e1000_read_phy_reg(&adapter->hw,
2202 			    PHY_1000T_CTRL, &phy_tmp);
2203 			if(phy_tmp & CR_1000T_MS_ENABLE) {
2204 				phy_tmp &= ~CR_1000T_MS_ENABLE;
2205 				e1000_write_phy_reg(&adapter->hw,
2206 				    PHY_1000T_CTRL, phy_tmp);
2207 				adapter->smartspeed++;
2208 				if(adapter->hw.mac.autoneg &&
2209 				   !e1000_copper_link_autoneg(&adapter->hw) &&
2210 				   !e1000_read_phy_reg(&adapter->hw,
2211 				    PHY_CONTROL, &phy_tmp)) {
2212 					phy_tmp |= (MII_CR_AUTO_NEG_EN |
2213 						    MII_CR_RESTART_AUTO_NEG);
2214 					e1000_write_phy_reg(&adapter->hw,
2215 					    PHY_CONTROL, phy_tmp);
2216 				}
2217 			}
2218 		}
2219 		return;
2220 	} else if(adapter->smartspeed == EM_SMARTSPEED_DOWNSHIFT) {
2221 		/* If still no link, perhaps using 2/3 pair cable */
2222 		e1000_read_phy_reg(&adapter->hw, PHY_1000T_CTRL, &phy_tmp);
2223 		phy_tmp |= CR_1000T_MS_ENABLE;
2224 		e1000_write_phy_reg(&adapter->hw, PHY_1000T_CTRL, phy_tmp);
2225 		if(adapter->hw.mac.autoneg &&
2226 		   !e1000_copper_link_autoneg(&adapter->hw) &&
2227 		   !e1000_read_phy_reg(&adapter->hw, PHY_CONTROL, &phy_tmp)) {
2228 			phy_tmp |= (MII_CR_AUTO_NEG_EN |
2229 				    MII_CR_RESTART_AUTO_NEG);
2230 			e1000_write_phy_reg(&adapter->hw, PHY_CONTROL, phy_tmp);
2231 		}
2232 	}
2233 	/* Restart process after EM_SMARTSPEED_MAX iterations */
2234 	if(adapter->smartspeed++ == EM_SMARTSPEED_MAX)
2235 		adapter->smartspeed = 0;
2236 }
2237 
2238 /*********************************************************************
2239  *
2240  *  Initialize the DMA Coalescing feature
2241  *
2242  **********************************************************************/
2243 static void
2244 igb_init_dmac(struct adapter *adapter, u32 pba)
2245 {
2246 	device_t	dev = adapter->dev;
2247 	struct e1000_hw *hw = &adapter->hw;
2248 	u32 		dmac, reg = ~E1000_DMACR_DMAC_EN;
2249 	u16		hwm;
2250 	u16		max_frame_size;
2251 
2252 	if (hw->mac.type == e1000_i211)
2253 		return;
2254 
2255 	max_frame_size = adapter->shared->isc_max_frame_size;
2256 	if (hw->mac.type > e1000_82580) {
2257 
2258 		if (adapter->dmac == 0) { /* Disabling it */
2259 			E1000_WRITE_REG(hw, E1000_DMACR, reg);
2260 			return;
2261 		} else
2262 			device_printf(dev, "DMA Coalescing enabled\n");
2263 
2264 		/* Set starting threshold */
2265 		E1000_WRITE_REG(hw, E1000_DMCTXTH, 0);
2266 
2267 		hwm = 64 * pba - max_frame_size / 16;
2268 		if (hwm < 64 * (pba - 6))
2269 			hwm = 64 * (pba - 6);
2270 		reg = E1000_READ_REG(hw, E1000_FCRTC);
2271 		reg &= ~E1000_FCRTC_RTH_COAL_MASK;
2272 		reg |= ((hwm << E1000_FCRTC_RTH_COAL_SHIFT)
2273 		    & E1000_FCRTC_RTH_COAL_MASK);
2274 		E1000_WRITE_REG(hw, E1000_FCRTC, reg);
2275 
2276 
2277 		dmac = pba - max_frame_size / 512;
2278 		if (dmac < pba - 10)
2279 			dmac = pba - 10;
2280 		reg = E1000_READ_REG(hw, E1000_DMACR);
2281 		reg &= ~E1000_DMACR_DMACTHR_MASK;
2282 		reg = ((dmac << E1000_DMACR_DMACTHR_SHIFT)
2283 		    & E1000_DMACR_DMACTHR_MASK);
2284 
2285 		/* transition to L0x or L1 if available..*/
2286 		reg |= (E1000_DMACR_DMAC_EN | E1000_DMACR_DMAC_LX_MASK);
2287 
2288 		/* Check if status is 2.5Gb backplane connection
2289 		* before configuration of watchdog timer, which is
2290 		* in msec values in 12.8usec intervals
2291 		* watchdog timer= msec values in 32usec intervals
2292 		* for non 2.5Gb connection
2293 		*/
2294 		if (hw->mac.type == e1000_i354) {
2295 			int status = E1000_READ_REG(hw, E1000_STATUS);
2296 			if ((status & E1000_STATUS_2P5_SKU) &&
2297 			    (!(status & E1000_STATUS_2P5_SKU_OVER)))
2298 				reg |= ((adapter->dmac * 5) >> 6);
2299 			else
2300 				reg |= (adapter->dmac >> 5);
2301 		} else {
2302 			reg |= (adapter->dmac >> 5);
2303 		}
2304 
2305 		E1000_WRITE_REG(hw, E1000_DMACR, reg);
2306 
2307 		E1000_WRITE_REG(hw, E1000_DMCRTRH, 0);
2308 
2309 		/* Set the interval before transition */
2310 		reg = E1000_READ_REG(hw, E1000_DMCTLX);
2311 		if (hw->mac.type == e1000_i350)
2312 			reg |= IGB_DMCTLX_DCFLUSH_DIS;
2313 		/*
2314 		** in 2.5Gb connection, TTLX unit is 0.4 usec
2315 		** which is 0x4*2 = 0xA. But delay is still 4 usec
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 |= 0xA;
2322 			else
2323 				reg |= 0x4;
2324 		} else {
2325 			reg |= 0x4;
2326 		}
2327 
2328 		E1000_WRITE_REG(hw, E1000_DMCTLX, reg);
2329 
2330 		/* free space in tx packet buffer to wake from DMA coal */
2331 		E1000_WRITE_REG(hw, E1000_DMCTXTH, (IGB_TXPBSIZE -
2332 		    (2 * max_frame_size)) >> 6);
2333 
2334 		/* make low power state decision controlled by DMA coal */
2335 		reg = E1000_READ_REG(hw, E1000_PCIEMISC);
2336 		reg &= ~E1000_PCIEMISC_LX_DECISION;
2337 		E1000_WRITE_REG(hw, E1000_PCIEMISC, reg);
2338 
2339 	} else if (hw->mac.type == e1000_82580) {
2340 		u32 reg = E1000_READ_REG(hw, E1000_PCIEMISC);
2341 		E1000_WRITE_REG(hw, E1000_PCIEMISC,
2342 		    reg & ~E1000_PCIEMISC_LX_DECISION);
2343 		E1000_WRITE_REG(hw, E1000_DMACR, 0);
2344 	}
2345 }
2346 
2347 static void
2348 em_reset(if_ctx_t ctx)
2349 {
2350 	device_t dev = iflib_get_dev(ctx);
2351 	struct adapter *adapter = iflib_get_softc(ctx);
2352 	struct ifnet *ifp = iflib_get_ifp(ctx);
2353 	struct e1000_hw *hw = &adapter->hw;
2354 	u16 rx_buffer_size;
2355 	u32 pba;
2356 
2357 	INIT_DEBUGOUT("em_reset: begin");
2358 	/* Let the firmware know the OS is in control */
2359 	em_get_hw_control(adapter);
2360 
2361 	/* Set up smart power down as default off on newer adapters. */
2362 	if (!em_smart_pwr_down && (hw->mac.type == e1000_82571 ||
2363 	    hw->mac.type == e1000_82572)) {
2364 		u16 phy_tmp = 0;
2365 
2366 		/* Speed up time to link by disabling smart power down. */
2367 		e1000_read_phy_reg(hw, IGP02E1000_PHY_POWER_MGMT, &phy_tmp);
2368 		phy_tmp &= ~IGP02E1000_PM_SPD;
2369 		e1000_write_phy_reg(hw, IGP02E1000_PHY_POWER_MGMT, phy_tmp);
2370 	}
2371 
2372 	/*
2373 	 * Packet Buffer Allocation (PBA)
2374 	 * Writing PBA sets the receive portion of the buffer
2375 	 * the remainder is used for the transmit buffer.
2376 	 */
2377 	switch (hw->mac.type) {
2378 	/* Total Packet Buffer on these is 48K */
2379 	case e1000_82571:
2380 	case e1000_82572:
2381 	case e1000_80003es2lan:
2382 			pba = E1000_PBA_32K; /* 32K for Rx, 16K for Tx */
2383 		break;
2384 	case e1000_82573: /* 82573: Total Packet Buffer is 32K */
2385 			pba = E1000_PBA_12K; /* 12K for Rx, 20K for Tx */
2386 		break;
2387 	case e1000_82574:
2388 	case e1000_82583:
2389 			pba = E1000_PBA_20K; /* 20K for Rx, 20K for Tx */
2390 		break;
2391 	case e1000_ich8lan:
2392 		pba = E1000_PBA_8K;
2393 		break;
2394 	case e1000_ich9lan:
2395 	case e1000_ich10lan:
2396 		/* Boost Receive side for jumbo frames */
2397 		if (adapter->hw.mac.max_frame_size > 4096)
2398 			pba = E1000_PBA_14K;
2399 		else
2400 			pba = E1000_PBA_10K;
2401 		break;
2402 	case e1000_pchlan:
2403 	case e1000_pch2lan:
2404 	case e1000_pch_lpt:
2405 	case e1000_pch_spt:
2406 		pba = E1000_PBA_26K;
2407 		break;
2408 	case e1000_82575:
2409 		pba = E1000_PBA_32K;
2410 		break;
2411 	case e1000_82576:
2412 	case e1000_vfadapt:
2413 		pba = E1000_READ_REG(hw, E1000_RXPBS);
2414 		pba &= E1000_RXPBS_SIZE_MASK_82576;
2415 		break;
2416 	case e1000_82580:
2417 	case e1000_i350:
2418 	case e1000_i354:
2419 	case e1000_vfadapt_i350:
2420 		pba = E1000_READ_REG(hw, E1000_RXPBS);
2421 		pba = e1000_rxpbs_adjust_82580(pba);
2422 		break;
2423 	case e1000_i210:
2424 	case e1000_i211:
2425 		pba = E1000_PBA_34K;
2426 		break;
2427 	default:
2428 		if (adapter->hw.mac.max_frame_size > 8192)
2429 			pba = E1000_PBA_40K; /* 40K for Rx, 24K for Tx */
2430 		else
2431 			pba = E1000_PBA_48K; /* 48K for Rx, 16K for Tx */
2432 	}
2433 
2434 	/* Special needs in case of Jumbo frames */
2435 	if ((hw->mac.type == e1000_82575) && (ifp->if_mtu > ETHERMTU)) {
2436 		u32 tx_space, min_tx, min_rx;
2437 		pba = E1000_READ_REG(hw, E1000_PBA);
2438 		tx_space = pba >> 16;
2439 		pba &= 0xffff;
2440 		min_tx = (adapter->hw.mac.max_frame_size +
2441 		    sizeof(struct e1000_tx_desc) - ETHERNET_FCS_SIZE) * 2;
2442 		min_tx = roundup2(min_tx, 1024);
2443 		min_tx >>= 10;
2444 		min_rx = adapter->hw.mac.max_frame_size;
2445 		min_rx = roundup2(min_rx, 1024);
2446 		min_rx >>= 10;
2447 		if (tx_space < min_tx &&
2448 		    ((min_tx - tx_space) < pba)) {
2449 			pba = pba - (min_tx - tx_space);
2450 			/*
2451 			 * if short on rx space, rx wins
2452 			 * and must trump tx adjustment
2453 			 */
2454 			if (pba < min_rx)
2455 				pba = min_rx;
2456 		}
2457 		E1000_WRITE_REG(hw, E1000_PBA, pba);
2458 	}
2459 
2460 	if (hw->mac.type < igb_mac_min)
2461 		E1000_WRITE_REG(&adapter->hw, E1000_PBA, pba);
2462 
2463 	INIT_DEBUGOUT1("em_reset: pba=%dK",pba);
2464 
2465 	/*
2466 	 * These parameters control the automatic generation (Tx) and
2467 	 * response (Rx) to Ethernet PAUSE frames.
2468 	 * - High water mark should allow for at least two frames to be
2469 	 *   received after sending an XOFF.
2470 	 * - Low water mark works best when it is very near the high water mark.
2471 	 *   This allows the receiver to restart by sending XON when it has
2472 	 *   drained a bit. Here we use an arbitrary value of 1500 which will
2473 	 *   restart after one full frame is pulled from the buffer. There
2474 	 *   could be several smaller frames in the buffer and if so they will
2475 	 *   not trigger the XON until their total number reduces the buffer
2476 	 *   by 1500.
2477 	 * - The pause time is fairly large at 1000 x 512ns = 512 usec.
2478 	 */
2479 	rx_buffer_size = (pba & 0xffff) << 10;
2480 	hw->fc.high_water = rx_buffer_size -
2481 	    roundup2(adapter->hw.mac.max_frame_size, 1024);
2482 	hw->fc.low_water = hw->fc.high_water - 1500;
2483 
2484 	if (adapter->fc) /* locally set flow control value? */
2485 		hw->fc.requested_mode = adapter->fc;
2486 	else
2487 		hw->fc.requested_mode = e1000_fc_full;
2488 
2489 	if (hw->mac.type == e1000_80003es2lan)
2490 		hw->fc.pause_time = 0xFFFF;
2491 	else
2492 		hw->fc.pause_time = EM_FC_PAUSE_TIME;
2493 
2494 	hw->fc.send_xon = TRUE;
2495 
2496 	/* Device specific overrides/settings */
2497 	switch (hw->mac.type) {
2498 	case e1000_pchlan:
2499 		/* Workaround: no TX flow ctrl for PCH */
2500 		hw->fc.requested_mode = e1000_fc_rx_pause;
2501 		hw->fc.pause_time = 0xFFFF; /* override */
2502 		if (if_getmtu(ifp) > ETHERMTU) {
2503 			hw->fc.high_water = 0x3500;
2504 			hw->fc.low_water = 0x1500;
2505 		} else {
2506 			hw->fc.high_water = 0x5000;
2507 			hw->fc.low_water = 0x3000;
2508 		}
2509 		hw->fc.refresh_time = 0x1000;
2510 		break;
2511 	case e1000_pch2lan:
2512 	case e1000_pch_lpt:
2513 	case e1000_pch_spt:
2514 		hw->fc.high_water = 0x5C20;
2515 		hw->fc.low_water = 0x5048;
2516 		hw->fc.pause_time = 0x0650;
2517 		hw->fc.refresh_time = 0x0400;
2518 		/* Jumbos need adjusted PBA */
2519 		if (if_getmtu(ifp) > ETHERMTU)
2520 			E1000_WRITE_REG(hw, E1000_PBA, 12);
2521 		else
2522 			E1000_WRITE_REG(hw, E1000_PBA, 26);
2523 		break;
2524 	case e1000_82575:
2525 	case e1000_82576:
2526 		/* 8-byte granularity */
2527 		hw->fc.low_water = hw->fc.high_water - 8;
2528 		break;
2529 	case e1000_82580:
2530 	case e1000_i350:
2531 	case e1000_i354:
2532 	case e1000_i210:
2533 	case e1000_i211:
2534 	case e1000_vfadapt:
2535 	case e1000_vfadapt_i350:
2536 		/* 16-byte granularity */
2537 		hw->fc.low_water = hw->fc.high_water - 16;
2538 		break;
2539 	case e1000_ich9lan:
2540 	case e1000_ich10lan:
2541 		if (if_getmtu(ifp) > ETHERMTU) {
2542 			hw->fc.high_water = 0x2800;
2543 			hw->fc.low_water = hw->fc.high_water - 8;
2544 			break;
2545 		}
2546 		/* FALLTHROUGH */
2547 	default:
2548 		if (hw->mac.type == e1000_80003es2lan)
2549 			hw->fc.pause_time = 0xFFFF;
2550 		break;
2551 	}
2552 
2553 	/* Issue a global reset */
2554 	e1000_reset_hw(hw);
2555 	if (adapter->hw.mac.type >= igb_mac_min) {
2556 		E1000_WRITE_REG(hw, E1000_WUC, 0);
2557 	} else {
2558 		E1000_WRITE_REG(hw, E1000_WUFC, 0);
2559 		em_disable_aspm(adapter);
2560 	}
2561 	if (adapter->flags & IGB_MEDIA_RESET) {
2562 		e1000_setup_init_funcs(hw, TRUE);
2563 		e1000_get_bus_info(hw);
2564 		adapter->flags &= ~IGB_MEDIA_RESET;
2565 	}
2566 	/* and a re-init */
2567 	if (e1000_init_hw(hw) < 0) {
2568 		device_printf(dev, "Hardware Initialization Failed\n");
2569 		return;
2570 	}
2571 	if (adapter->hw.mac.type >= igb_mac_min)
2572 		igb_init_dmac(adapter, pba);
2573 
2574 	E1000_WRITE_REG(hw, E1000_VET, ETHERTYPE_VLAN);
2575 	e1000_get_phy_info(hw);
2576 	e1000_check_for_link(hw);
2577 }
2578 
2579 #define RSSKEYLEN 10
2580 static void
2581 em_initialize_rss_mapping(struct adapter *adapter)
2582 {
2583 	uint8_t  rss_key[4 * RSSKEYLEN];
2584 	uint32_t reta = 0;
2585 	struct e1000_hw	*hw = &adapter->hw;
2586 	int i;
2587 
2588 	/*
2589 	 * Configure RSS key
2590 	 */
2591 	arc4rand(rss_key, sizeof(rss_key), 0);
2592 	for (i = 0; i < RSSKEYLEN; ++i) {
2593 		uint32_t rssrk = 0;
2594 
2595 		rssrk = EM_RSSRK_VAL(rss_key, i);
2596 		E1000_WRITE_REG(hw,E1000_RSSRK(i), rssrk);
2597 	}
2598 
2599 	/*
2600 	 * Configure RSS redirect table in following fashion:
2601 	 * (hash & ring_cnt_mask) == rdr_table[(hash & rdr_table_mask)]
2602 	 */
2603 	for (i = 0; i < sizeof(reta); ++i) {
2604 		uint32_t q;
2605 
2606 		q = (i % adapter->rx_num_queues) << 7;
2607 		reta |= q << (8 * i);
2608 	}
2609 
2610 	for (i = 0; i < 32; ++i)
2611 		E1000_WRITE_REG(hw, E1000_RETA(i), reta);
2612 
2613 	E1000_WRITE_REG(hw, E1000_MRQC, E1000_MRQC_RSS_ENABLE_2Q |
2614 			E1000_MRQC_RSS_FIELD_IPV4_TCP |
2615 			E1000_MRQC_RSS_FIELD_IPV4 |
2616 			E1000_MRQC_RSS_FIELD_IPV6_TCP_EX |
2617 			E1000_MRQC_RSS_FIELD_IPV6_EX |
2618 			E1000_MRQC_RSS_FIELD_IPV6);
2619 
2620 }
2621 
2622 static void
2623 igb_initialize_rss_mapping(struct adapter *adapter)
2624 {
2625 	struct e1000_hw *hw = &adapter->hw;
2626 	int i;
2627 	int queue_id;
2628 	u32 reta;
2629 	u32 rss_key[10], mrqc, shift = 0;
2630 
2631 	/* XXX? */
2632 	if (adapter->hw.mac.type == e1000_82575)
2633 		shift = 6;
2634 
2635 	/*
2636 	 * The redirection table controls which destination
2637 	 * queue each bucket redirects traffic to.
2638 	 * Each DWORD represents four queues, with the LSB
2639 	 * being the first queue in the DWORD.
2640 	 *
2641 	 * This just allocates buckets to queues using round-robin
2642 	 * allocation.
2643 	 *
2644 	 * NOTE: It Just Happens to line up with the default
2645 	 * RSS allocation method.
2646 	 */
2647 
2648 	/* Warning FM follows */
2649 	reta = 0;
2650 	for (i = 0; i < 128; i++) {
2651 #ifdef RSS
2652 		queue_id = rss_get_indirection_to_bucket(i);
2653 		/*
2654 		 * If we have more queues than buckets, we'll
2655 		 * end up mapping buckets to a subset of the
2656 		 * queues.
2657 		 *
2658 		 * If we have more buckets than queues, we'll
2659 		 * end up instead assigning multiple buckets
2660 		 * to queues.
2661 		 *
2662 		 * Both are suboptimal, but we need to handle
2663 		 * the case so we don't go out of bounds
2664 		 * indexing arrays and such.
2665 		 */
2666 		queue_id = queue_id % adapter->rx_num_queues;
2667 #else
2668 		queue_id = (i % adapter->rx_num_queues);
2669 #endif
2670 		/* Adjust if required */
2671 		queue_id = queue_id << shift;
2672 
2673 		/*
2674 		 * The low 8 bits are for hash value (n+0);
2675 		 * The next 8 bits are for hash value (n+1), etc.
2676 		 */
2677 		reta = reta >> 8;
2678 		reta = reta | ( ((uint32_t) queue_id) << 24);
2679 		if ((i & 3) == 3) {
2680 			E1000_WRITE_REG(hw, E1000_RETA(i >> 2), reta);
2681 			reta = 0;
2682 		}
2683 	}
2684 
2685 	/* Now fill in hash table */
2686 
2687 	/*
2688 	 * MRQC: Multiple Receive Queues Command
2689 	 * Set queuing to RSS control, number depends on the device.
2690 	 */
2691 	mrqc = E1000_MRQC_ENABLE_RSS_8Q;
2692 
2693 #ifdef RSS
2694 	/* XXX ew typecasting */
2695 	rss_getkey((uint8_t *) &rss_key);
2696 #else
2697 	arc4rand(&rss_key, sizeof(rss_key), 0);
2698 #endif
2699 	for (i = 0; i < 10; i++)
2700 		E1000_WRITE_REG_ARRAY(hw, E1000_RSSRK(0), i, rss_key[i]);
2701 
2702 	/*
2703 	 * Configure the RSS fields to hash upon.
2704 	 */
2705 	mrqc |= (E1000_MRQC_RSS_FIELD_IPV4 |
2706 	    E1000_MRQC_RSS_FIELD_IPV4_TCP);
2707 	mrqc |= (E1000_MRQC_RSS_FIELD_IPV6 |
2708 	    E1000_MRQC_RSS_FIELD_IPV6_TCP);
2709 	mrqc |=( E1000_MRQC_RSS_FIELD_IPV4_UDP |
2710 	    E1000_MRQC_RSS_FIELD_IPV6_UDP);
2711 	mrqc |=( E1000_MRQC_RSS_FIELD_IPV6_UDP_EX |
2712 	    E1000_MRQC_RSS_FIELD_IPV6_TCP_EX);
2713 
2714 	E1000_WRITE_REG(hw, E1000_MRQC, mrqc);
2715 }
2716 
2717 /*********************************************************************
2718  *
2719  *  Setup networking device structure and register an interface.
2720  *
2721  **********************************************************************/
2722 static int
2723 em_setup_interface(if_ctx_t ctx)
2724 {
2725 	struct ifnet *ifp = iflib_get_ifp(ctx);
2726 	struct adapter *adapter = iflib_get_softc(ctx);
2727 	if_softc_ctx_t scctx = adapter->shared;
2728 	uint64_t cap = 0;
2729 
2730 	INIT_DEBUGOUT("em_setup_interface: begin");
2731 
2732 	/* TSO parameters */
2733 	if_sethwtsomax(ifp, IP_MAXPACKET);
2734 	/* Take m_pullup(9)'s in em_xmit() w/ TSO into acount. */
2735 	if_sethwtsomaxsegcount(ifp, EM_MAX_SCATTER - 5);
2736 	if_sethwtsomaxsegsize(ifp, EM_TSO_SEG_SIZE);
2737 
2738 	/* Single Queue */
2739 	if (adapter->tx_num_queues == 1) {
2740 		if_setsendqlen(ifp, scctx->isc_ntxd[0] - 1);
2741 		if_setsendqready(ifp);
2742 	}
2743 
2744 	cap = IFCAP_HWCSUM | IFCAP_VLAN_HWCSUM | IFCAP_TSO4;
2745 	cap |= IFCAP_VLAN_HWTAGGING | IFCAP_VLAN_HWTSO | IFCAP_VLAN_MTU;
2746 
2747 	/*
2748 	 * Tell the upper layer(s) we
2749 	 * support full VLAN capability
2750 	 */
2751 	if_setifheaderlen(ifp, sizeof(struct ether_vlan_header));
2752 	if_setcapabilitiesbit(ifp, cap, 0);
2753 
2754 	/*
2755 	 * Don't turn this on by default, if vlans are
2756 	 * created on another pseudo device (eg. lagg)
2757 	 * then vlan events are not passed thru, breaking
2758 	 * operation, but with HW FILTER off it works. If
2759 	 * using vlans directly on the em driver you can
2760 	 * enable this and get full hardware tag filtering.
2761 	 */
2762 	if_setcapabilitiesbit(ifp, IFCAP_VLAN_HWFILTER,0);
2763 
2764 	/* Enable only WOL MAGIC by default */
2765 	if (adapter->wol) {
2766 		if_setcapenablebit(ifp, IFCAP_WOL_MAGIC,
2767 			    IFCAP_WOL_MCAST| IFCAP_WOL_UCAST);
2768 	} else {
2769 		if_setcapenablebit(ifp, 0, IFCAP_WOL_MAGIC |
2770 			     IFCAP_WOL_MCAST| IFCAP_WOL_UCAST);
2771 	}
2772 
2773 	/*
2774 	 * Specify the media types supported by this adapter and register
2775 	 * callbacks to update media and link information
2776 	 */
2777 	if ((adapter->hw.phy.media_type == e1000_media_type_fiber) ||
2778 	    (adapter->hw.phy.media_type == e1000_media_type_internal_serdes)) {
2779 		u_char fiber_type = IFM_1000_SX;	/* default type */
2780 
2781 		if (adapter->hw.mac.type == e1000_82545)
2782 			fiber_type = IFM_1000_LX;
2783 		ifmedia_add(adapter->media, IFM_ETHER | fiber_type | IFM_FDX, 0, NULL);
2784 		ifmedia_add(adapter->media, IFM_ETHER | fiber_type, 0, NULL);
2785 	} else {
2786 		ifmedia_add(adapter->media, IFM_ETHER | IFM_10_T, 0, NULL);
2787 		ifmedia_add(adapter->media, IFM_ETHER | IFM_10_T | IFM_FDX, 0, NULL);
2788 		ifmedia_add(adapter->media, IFM_ETHER | IFM_100_TX, 0, NULL);
2789 		ifmedia_add(adapter->media, IFM_ETHER | IFM_100_TX | IFM_FDX, 0, NULL);
2790 		if (adapter->hw.phy.type != e1000_phy_ife) {
2791 			ifmedia_add(adapter->media, IFM_ETHER | IFM_1000_T | IFM_FDX, 0, NULL);
2792 			ifmedia_add(adapter->media, IFM_ETHER | IFM_1000_T, 0, NULL);
2793 		}
2794 	}
2795 	ifmedia_add(adapter->media, IFM_ETHER | IFM_AUTO, 0, NULL);
2796 	ifmedia_set(adapter->media, IFM_ETHER | IFM_AUTO);
2797 	return (0);
2798 }
2799 
2800 static int
2801 em_if_tx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int ntxqs, int ntxqsets)
2802 {
2803 	struct adapter *adapter = iflib_get_softc(ctx);
2804 	if_softc_ctx_t scctx = adapter->shared;
2805 	int error = E1000_SUCCESS;
2806 	struct em_tx_queue *que;
2807 	int i, j;
2808 
2809 	MPASS(adapter->tx_num_queues > 0);
2810 	MPASS(adapter->tx_num_queues == ntxqsets);
2811 
2812 	/* First allocate the top level queue structs */
2813 	if (!(adapter->tx_queues =
2814 	    (struct em_tx_queue *) malloc(sizeof(struct em_tx_queue) *
2815 	    adapter->tx_num_queues, M_DEVBUF, M_NOWAIT | M_ZERO))) {
2816 		device_printf(iflib_get_dev(ctx), "Unable to allocate queue memory\n");
2817 		return(ENOMEM);
2818 	}
2819 
2820 	for (i = 0, que = adapter->tx_queues; i < adapter->tx_num_queues; i++, que++) {
2821 		/* Set up some basics */
2822 
2823 		struct tx_ring *txr = &que->txr;
2824 		txr->adapter = que->adapter = adapter;
2825 		que->me = txr->me =  i;
2826 
2827 		/* Allocate report status array */
2828 		if (!(txr->tx_rsq = (qidx_t *) malloc(sizeof(qidx_t) * scctx->isc_ntxd[0], M_DEVBUF, M_NOWAIT | M_ZERO))) {
2829 			device_printf(iflib_get_dev(ctx), "failed to allocate rs_idxs memory\n");
2830 			error = ENOMEM;
2831 			goto fail;
2832 		}
2833 		for (j = 0; j < scctx->isc_ntxd[0]; j++)
2834 			txr->tx_rsq[j] = QIDX_INVALID;
2835 		/* get the virtual and physical address of the hardware queues */
2836 		txr->tx_base = (struct e1000_tx_desc *)vaddrs[i*ntxqs];
2837 		txr->tx_paddr = paddrs[i*ntxqs];
2838 	}
2839 
2840 	device_printf(iflib_get_dev(ctx), "allocated for %d tx_queues\n", adapter->tx_num_queues);
2841 	return (0);
2842 fail:
2843 	em_if_queues_free(ctx);
2844 	return (error);
2845 }
2846 
2847 static int
2848 em_if_rx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int nrxqs, int nrxqsets)
2849 {
2850 	struct adapter *adapter = iflib_get_softc(ctx);
2851 	int error = E1000_SUCCESS;
2852 	struct em_rx_queue *que;
2853 	int i;
2854 
2855 	MPASS(adapter->rx_num_queues > 0);
2856 	MPASS(adapter->rx_num_queues == nrxqsets);
2857 
2858 	/* First allocate the top level queue structs */
2859 	if (!(adapter->rx_queues =
2860 	    (struct em_rx_queue *) malloc(sizeof(struct em_rx_queue) *
2861 	    adapter->rx_num_queues, M_DEVBUF, M_NOWAIT | M_ZERO))) {
2862 		device_printf(iflib_get_dev(ctx), "Unable to allocate queue memory\n");
2863 		error = ENOMEM;
2864 		goto fail;
2865 	}
2866 
2867 	for (i = 0, que = adapter->rx_queues; i < nrxqsets; i++, que++) {
2868 		/* Set up some basics */
2869 		struct rx_ring *rxr = &que->rxr;
2870 		rxr->adapter = que->adapter = adapter;
2871 		rxr->que = que;
2872 		que->me = rxr->me =  i;
2873 
2874 		/* get the virtual and physical address of the hardware queues */
2875 		rxr->rx_base = (union e1000_rx_desc_extended *)vaddrs[i*nrxqs];
2876 		rxr->rx_paddr = paddrs[i*nrxqs];
2877 	}
2878 
2879 	device_printf(iflib_get_dev(ctx), "allocated for %d rx_queues\n", adapter->rx_num_queues);
2880 
2881 	return (0);
2882 fail:
2883 	em_if_queues_free(ctx);
2884 	return (error);
2885 }
2886 
2887 static void
2888 em_if_queues_free(if_ctx_t ctx)
2889 {
2890 	struct adapter *adapter = iflib_get_softc(ctx);
2891 	struct em_tx_queue *tx_que = adapter->tx_queues;
2892 	struct em_rx_queue *rx_que = adapter->rx_queues;
2893 
2894 	if (tx_que != NULL) {
2895 		for (int i = 0; i < adapter->tx_num_queues; i++, tx_que++) {
2896 			struct tx_ring *txr = &tx_que->txr;
2897 			if (txr->tx_rsq == NULL)
2898 				break;
2899 
2900 			free(txr->tx_rsq, M_DEVBUF);
2901 			txr->tx_rsq = NULL;
2902 		}
2903 		free(adapter->tx_queues, M_DEVBUF);
2904 		adapter->tx_queues = NULL;
2905 	}
2906 
2907 	if (rx_que != NULL) {
2908 		free(adapter->rx_queues, M_DEVBUF);
2909 		adapter->rx_queues = NULL;
2910 	}
2911 
2912 	em_release_hw_control(adapter);
2913 
2914 	if (adapter->mta != NULL) {
2915 		free(adapter->mta, M_DEVBUF);
2916 	}
2917 }
2918 
2919 /*********************************************************************
2920  *
2921  *  Enable transmit unit.
2922  *
2923  **********************************************************************/
2924 static void
2925 em_initialize_transmit_unit(if_ctx_t ctx)
2926 {
2927 	struct adapter *adapter = iflib_get_softc(ctx);
2928 	if_softc_ctx_t scctx = adapter->shared;
2929 	struct em_tx_queue *que;
2930 	struct tx_ring	*txr;
2931 	struct e1000_hw	*hw = &adapter->hw;
2932 	u32 tctl, txdctl = 0, tarc, tipg = 0;
2933 
2934 	INIT_DEBUGOUT("em_initialize_transmit_unit: begin");
2935 
2936 	for (int i = 0; i < adapter->tx_num_queues; i++, txr++) {
2937 		u64 bus_addr;
2938 		caddr_t offp, endp;
2939 
2940 		que = &adapter->tx_queues[i];
2941 		txr = &que->txr;
2942 		bus_addr = txr->tx_paddr;
2943 
2944 		/* Clear checksum offload context. */
2945 		offp = (caddr_t)&txr->csum_flags;
2946 		endp = (caddr_t)(txr + 1);
2947 		bzero(offp, endp - offp);
2948 
2949 		/* Base and Len of TX Ring */
2950 		E1000_WRITE_REG(hw, E1000_TDLEN(i),
2951 		    scctx->isc_ntxd[0] * sizeof(struct e1000_tx_desc));
2952 		E1000_WRITE_REG(hw, E1000_TDBAH(i),
2953 		    (u32)(bus_addr >> 32));
2954 		E1000_WRITE_REG(hw, E1000_TDBAL(i),
2955 		    (u32)bus_addr);
2956 		/* Init the HEAD/TAIL indices */
2957 		E1000_WRITE_REG(hw, E1000_TDT(i), 0);
2958 		E1000_WRITE_REG(hw, E1000_TDH(i), 0);
2959 
2960 		HW_DEBUGOUT2("Base = %x, Length = %x\n",
2961 		    E1000_READ_REG(&adapter->hw, E1000_TDBAL(i)),
2962 		    E1000_READ_REG(&adapter->hw, E1000_TDLEN(i)));
2963 
2964 		txdctl = 0; /* clear txdctl */
2965 		txdctl |= 0x1f; /* PTHRESH */
2966 		txdctl |= 1 << 8; /* HTHRESH */
2967 		txdctl |= 1 << 16;/* WTHRESH */
2968 		txdctl |= 1 << 22; /* Reserved bit 22 must always be 1 */
2969 		txdctl |= E1000_TXDCTL_GRAN;
2970 		txdctl |= 1 << 25; /* LWTHRESH */
2971 
2972 		E1000_WRITE_REG(hw, E1000_TXDCTL(i), txdctl);
2973 	}
2974 
2975 	/* Set the default values for the Tx Inter Packet Gap timer */
2976 	switch (adapter->hw.mac.type) {
2977 	case e1000_80003es2lan:
2978 		tipg = DEFAULT_82543_TIPG_IPGR1;
2979 		tipg |= DEFAULT_80003ES2LAN_TIPG_IPGR2 <<
2980 		    E1000_TIPG_IPGR2_SHIFT;
2981 		break;
2982 	case e1000_82542:
2983 		tipg = DEFAULT_82542_TIPG_IPGT;
2984 		tipg |= DEFAULT_82542_TIPG_IPGR1 << E1000_TIPG_IPGR1_SHIFT;
2985 		tipg |= DEFAULT_82542_TIPG_IPGR2 << E1000_TIPG_IPGR2_SHIFT;
2986 		break;
2987 	default:
2988 		if ((adapter->hw.phy.media_type == e1000_media_type_fiber) ||
2989 		    (adapter->hw.phy.media_type ==
2990 		    e1000_media_type_internal_serdes))
2991 			tipg = DEFAULT_82543_TIPG_IPGT_FIBER;
2992 		else
2993 			tipg = DEFAULT_82543_TIPG_IPGT_COPPER;
2994 		tipg |= DEFAULT_82543_TIPG_IPGR1 << E1000_TIPG_IPGR1_SHIFT;
2995 		tipg |= DEFAULT_82543_TIPG_IPGR2 << E1000_TIPG_IPGR2_SHIFT;
2996 	}
2997 
2998 	E1000_WRITE_REG(&adapter->hw, E1000_TIPG, tipg);
2999 	E1000_WRITE_REG(&adapter->hw, E1000_TIDV, adapter->tx_int_delay.value);
3000 
3001 	if(adapter->hw.mac.type >= e1000_82540)
3002 		E1000_WRITE_REG(&adapter->hw, E1000_TADV,
3003 		    adapter->tx_abs_int_delay.value);
3004 
3005 	if ((adapter->hw.mac.type == e1000_82571) ||
3006 	    (adapter->hw.mac.type == e1000_82572)) {
3007 		tarc = E1000_READ_REG(&adapter->hw, E1000_TARC(0));
3008 		tarc |= TARC_SPEED_MODE_BIT;
3009 		E1000_WRITE_REG(&adapter->hw, E1000_TARC(0), tarc);
3010 	} else if (adapter->hw.mac.type == e1000_80003es2lan) {
3011 		/* errata: program both queues to unweighted RR */
3012 		tarc = E1000_READ_REG(&adapter->hw, E1000_TARC(0));
3013 		tarc |= 1;
3014 		E1000_WRITE_REG(&adapter->hw, E1000_TARC(0), tarc);
3015 		tarc = E1000_READ_REG(&adapter->hw, E1000_TARC(1));
3016 		tarc |= 1;
3017 		E1000_WRITE_REG(&adapter->hw, E1000_TARC(1), tarc);
3018 	} else if (adapter->hw.mac.type == e1000_82574) {
3019 		tarc = E1000_READ_REG(&adapter->hw, E1000_TARC(0));
3020 		tarc |= TARC_ERRATA_BIT;
3021 		if ( adapter->tx_num_queues > 1) {
3022 			tarc |= (TARC_COMPENSATION_MODE | TARC_MQ_FIX);
3023 			E1000_WRITE_REG(&adapter->hw, E1000_TARC(0), tarc);
3024 			E1000_WRITE_REG(&adapter->hw, E1000_TARC(1), tarc);
3025 		} else
3026 			E1000_WRITE_REG(&adapter->hw, E1000_TARC(0), tarc);
3027 	}
3028 
3029 	if (adapter->tx_int_delay.value > 0)
3030 		adapter->txd_cmd |= E1000_TXD_CMD_IDE;
3031 
3032 	/* Program the Transmit Control Register */
3033 	tctl = E1000_READ_REG(&adapter->hw, E1000_TCTL);
3034 	tctl &= ~E1000_TCTL_CT;
3035 	tctl |= (E1000_TCTL_PSP | E1000_TCTL_RTLC | E1000_TCTL_EN |
3036 		   (E1000_COLLISION_THRESHOLD << E1000_CT_SHIFT));
3037 
3038 	if (adapter->hw.mac.type >= e1000_82571)
3039 		tctl |= E1000_TCTL_MULR;
3040 
3041 	/* This write will effectively turn on the transmit unit. */
3042 	E1000_WRITE_REG(&adapter->hw, E1000_TCTL, tctl);
3043 
3044 	if (hw->mac.type == e1000_pch_spt) {
3045 		u32 reg;
3046 		reg = E1000_READ_REG(hw, E1000_IOSFPC);
3047 		reg |= E1000_RCTL_RDMTS_HEX;
3048 		E1000_WRITE_REG(hw, E1000_IOSFPC, reg);
3049 		reg = E1000_READ_REG(hw, E1000_TARC(0));
3050 		reg |= E1000_TARC0_CB_MULTIQ_3_REQ;
3051 		E1000_WRITE_REG(hw, E1000_TARC(0), reg);
3052 	}
3053 }
3054 
3055 /*********************************************************************
3056  *
3057  *  Enable receive unit.
3058  *
3059  **********************************************************************/
3060 
3061 static void
3062 em_initialize_receive_unit(if_ctx_t ctx)
3063 {
3064 	struct adapter *adapter = iflib_get_softc(ctx);
3065 	if_softc_ctx_t scctx = adapter->shared;
3066 	struct ifnet *ifp = iflib_get_ifp(ctx);
3067 	struct e1000_hw	*hw = &adapter->hw;
3068 	struct em_rx_queue *que;
3069 	int i;
3070 	u32 rctl, rxcsum, rfctl;
3071 
3072 	INIT_DEBUGOUT("em_initialize_receive_units: begin");
3073 
3074 	/*
3075 	 * Make sure receives are disabled while setting
3076 	 * up the descriptor ring
3077 	 */
3078 	rctl = E1000_READ_REG(hw, E1000_RCTL);
3079 	/* Do not disable if ever enabled on this hardware */
3080 	if ((hw->mac.type != e1000_82574) && (hw->mac.type != e1000_82583))
3081 		E1000_WRITE_REG(hw, E1000_RCTL, rctl & ~E1000_RCTL_EN);
3082 
3083 	/* Setup the Receive Control Register */
3084 	rctl &= ~(3 << E1000_RCTL_MO_SHIFT);
3085 	rctl |= E1000_RCTL_EN | E1000_RCTL_BAM |
3086 	    E1000_RCTL_LBM_NO | E1000_RCTL_RDMTS_HALF |
3087 	    (hw->mac.mc_filter_type << E1000_RCTL_MO_SHIFT);
3088 
3089 	/* Do not store bad packets */
3090 	rctl &= ~E1000_RCTL_SBP;
3091 
3092 	/* Enable Long Packet receive */
3093 	if (if_getmtu(ifp) > ETHERMTU)
3094 		rctl |= E1000_RCTL_LPE;
3095 	else
3096 		rctl &= ~E1000_RCTL_LPE;
3097 
3098 	/* Strip the CRC */
3099 	if (!em_disable_crc_stripping)
3100 		rctl |= E1000_RCTL_SECRC;
3101 
3102 	if (adapter->hw.mac.type >= e1000_82540) {
3103 		E1000_WRITE_REG(&adapter->hw, E1000_RADV,
3104 			    adapter->rx_abs_int_delay.value);
3105 
3106 		/*
3107 		 * Set the interrupt throttling rate. Value is calculated
3108 		 * as DEFAULT_ITR = 1/(MAX_INTS_PER_SEC * 256ns)
3109 		 */
3110 		E1000_WRITE_REG(hw, E1000_ITR, DEFAULT_ITR);
3111 	}
3112 	E1000_WRITE_REG(&adapter->hw, E1000_RDTR,
3113 	    adapter->rx_int_delay.value);
3114 
3115 	/* Use extended rx descriptor formats */
3116 	rfctl = E1000_READ_REG(hw, E1000_RFCTL);
3117 	rfctl |= E1000_RFCTL_EXTEN;
3118 	/*
3119 	 * When using MSIX interrupts we need to throttle
3120 	 * using the EITR register (82574 only)
3121 	 */
3122 	if (hw->mac.type == e1000_82574) {
3123 		for (int i = 0; i < 4; i++)
3124 			E1000_WRITE_REG(hw, E1000_EITR_82574(i),
3125 			    DEFAULT_ITR);
3126 		/* Disable accelerated acknowledge */
3127 		rfctl |= E1000_RFCTL_ACK_DIS;
3128 	}
3129 	E1000_WRITE_REG(hw, E1000_RFCTL, rfctl);
3130 
3131 	rxcsum = E1000_READ_REG(hw, E1000_RXCSUM);
3132 	if (if_getcapenable(ifp) & IFCAP_RXCSUM &&
3133 	    adapter->hw.mac.type >= e1000_82543) {
3134 		if (adapter->tx_num_queues > 1) {
3135 			if (adapter->hw.mac.type >= igb_mac_min) {
3136 				rxcsum |= E1000_RXCSUM_PCSD;
3137 				if (hw->mac.type != e1000_82575)
3138 					rxcsum |= E1000_RXCSUM_CRCOFL;
3139 			} else
3140 				rxcsum |= E1000_RXCSUM_TUOFL |
3141 					E1000_RXCSUM_IPOFL |
3142 					E1000_RXCSUM_PCSD;
3143 		} else {
3144 			if (adapter->hw.mac.type >= igb_mac_min)
3145 				rxcsum |= E1000_RXCSUM_IPPCSE;
3146 			else
3147 				rxcsum |= E1000_RXCSUM_TUOFL | E1000_RXCSUM_IPOFL;
3148 			if (adapter->hw.mac.type > e1000_82575)
3149 				rxcsum |= E1000_RXCSUM_CRCOFL;
3150 		}
3151 	} else
3152 		rxcsum &= ~E1000_RXCSUM_TUOFL;
3153 
3154 	E1000_WRITE_REG(hw, E1000_RXCSUM, rxcsum);
3155 
3156 	if (adapter->rx_num_queues > 1) {
3157 		if (adapter->hw.mac.type >= igb_mac_min)
3158 			igb_initialize_rss_mapping(adapter);
3159 		else
3160 			em_initialize_rss_mapping(adapter);
3161 	}
3162 
3163 	/*
3164 	 * XXX TEMPORARY WORKAROUND: on some systems with 82573
3165 	 * long latencies are observed, like Lenovo X60. This
3166 	 * change eliminates the problem, but since having positive
3167 	 * values in RDTR is a known source of problems on other
3168 	 * platforms another solution is being sought.
3169 	 */
3170 	if (hw->mac.type == e1000_82573)
3171 		E1000_WRITE_REG(hw, E1000_RDTR, 0x20);
3172 
3173 	for (i = 0, que = adapter->rx_queues; i < adapter->rx_num_queues; i++, que++) {
3174 		struct rx_ring *rxr = &que->rxr;
3175 		/* Setup the Base and Length of the Rx Descriptor Ring */
3176 		u64 bus_addr = rxr->rx_paddr;
3177 #if 0
3178 		u32 rdt = adapter->rx_num_queues -1;  /* default */
3179 #endif
3180 
3181 		E1000_WRITE_REG(hw, E1000_RDLEN(i),
3182 		    scctx->isc_nrxd[0] * sizeof(union e1000_rx_desc_extended));
3183 		E1000_WRITE_REG(hw, E1000_RDBAH(i), (u32)(bus_addr >> 32));
3184 		E1000_WRITE_REG(hw, E1000_RDBAL(i), (u32)bus_addr);
3185 		/* Setup the Head and Tail Descriptor Pointers */
3186 		E1000_WRITE_REG(hw, E1000_RDH(i), 0);
3187 		E1000_WRITE_REG(hw, E1000_RDT(i), 0);
3188 	}
3189 
3190 	/*
3191 	 * Set PTHRESH for improved jumbo performance
3192 	 * According to 10.2.5.11 of Intel 82574 Datasheet,
3193 	 * RXDCTL(1) is written whenever RXDCTL(0) is written.
3194 	 * Only write to RXDCTL(1) if there is a need for different
3195 	 * settings.
3196 	 */
3197 
3198 	if (((adapter->hw.mac.type == e1000_ich9lan) ||
3199 	    (adapter->hw.mac.type == e1000_pch2lan) ||
3200 	    (adapter->hw.mac.type == e1000_ich10lan)) &&
3201 	    (if_getmtu(ifp) > ETHERMTU)) {
3202 		u32 rxdctl = E1000_READ_REG(hw, E1000_RXDCTL(0));
3203 		E1000_WRITE_REG(hw, E1000_RXDCTL(0), rxdctl | 3);
3204 	} else if (adapter->hw.mac.type == e1000_82574) {
3205 		for (int i = 0; i < adapter->rx_num_queues; i++) {
3206 			u32 rxdctl = E1000_READ_REG(hw, E1000_RXDCTL(i));
3207 			rxdctl |= 0x20; /* PTHRESH */
3208 			rxdctl |= 4 << 8; /* HTHRESH */
3209 			rxdctl |= 4 << 16;/* WTHRESH */
3210 			rxdctl |= 1 << 24; /* Switch to granularity */
3211 			E1000_WRITE_REG(hw, E1000_RXDCTL(i), rxdctl);
3212 		}
3213 	} else if (adapter->hw.mac.type >= igb_mac_min) {
3214 		u32 psize, srrctl = 0;
3215 
3216 		if (if_getmtu(ifp) > ETHERMTU) {
3217 			/* Set maximum packet len */
3218 			if (adapter->rx_mbuf_sz <= 4096) {
3219 				srrctl |= 4096 >> E1000_SRRCTL_BSIZEPKT_SHIFT;
3220 				rctl |= E1000_RCTL_SZ_4096 | E1000_RCTL_BSEX;
3221 			} else if (adapter->rx_mbuf_sz > 4096) {
3222 				srrctl |= 8192 >> E1000_SRRCTL_BSIZEPKT_SHIFT;
3223 				rctl |= E1000_RCTL_SZ_8192 | E1000_RCTL_BSEX;
3224 			}
3225 			psize = scctx->isc_max_frame_size;
3226 			/* are we on a vlan? */
3227 			if (ifp->if_vlantrunk != NULL)
3228 				psize += VLAN_TAG_SIZE;
3229 			E1000_WRITE_REG(&adapter->hw, E1000_RLPML, psize);
3230 		} else {
3231 			srrctl |= 2048 >> E1000_SRRCTL_BSIZEPKT_SHIFT;
3232 			rctl |= E1000_RCTL_SZ_2048;
3233 		}
3234 
3235 		/*
3236 		 * If TX flow control is disabled and there's >1 queue defined,
3237 		 * enable DROP.
3238 		 *
3239 		 * This drops frames rather than hanging the RX MAC for all queues.
3240 		 */
3241 		if ((adapter->rx_num_queues > 1) &&
3242 		    (adapter->fc == e1000_fc_none ||
3243 		     adapter->fc == e1000_fc_rx_pause)) {
3244 			srrctl |= E1000_SRRCTL_DROP_EN;
3245 		}
3246 			/* Setup the Base and Length of the Rx Descriptor Rings */
3247 		for (i = 0, que = adapter->rx_queues; i < adapter->rx_num_queues; i++, que++) {
3248 			struct rx_ring *rxr = &que->rxr;
3249 			u64 bus_addr = rxr->rx_paddr;
3250 			u32 rxdctl;
3251 
3252 #ifdef notyet
3253 			/* Configure for header split? -- ignore for now */
3254 			rxr->hdr_split = igb_header_split;
3255 #else
3256 			srrctl |= E1000_SRRCTL_DESCTYPE_ADV_ONEBUF;
3257 #endif
3258 
3259 			E1000_WRITE_REG(hw, E1000_RDLEN(i),
3260 					scctx->isc_nrxd[0] * sizeof(struct e1000_rx_desc));
3261 			E1000_WRITE_REG(hw, E1000_RDBAH(i),
3262 					(uint32_t)(bus_addr >> 32));
3263 			E1000_WRITE_REG(hw, E1000_RDBAL(i),
3264 					(uint32_t)bus_addr);
3265 			E1000_WRITE_REG(hw, E1000_SRRCTL(i), srrctl);
3266 			/* Enable this Queue */
3267 			rxdctl = E1000_READ_REG(hw, E1000_RXDCTL(i));
3268 			rxdctl |= E1000_RXDCTL_QUEUE_ENABLE;
3269 			rxdctl &= 0xFFF00000;
3270 			rxdctl |= IGB_RX_PTHRESH;
3271 			rxdctl |= IGB_RX_HTHRESH << 8;
3272 			rxdctl |= IGB_RX_WTHRESH << 16;
3273 			E1000_WRITE_REG(hw, E1000_RXDCTL(i), rxdctl);
3274 		}
3275 	} else if (adapter->hw.mac.type >= e1000_pch2lan) {
3276 		if (if_getmtu(ifp) > ETHERMTU)
3277 			e1000_lv_jumbo_workaround_ich8lan(hw, TRUE);
3278 		else
3279 			e1000_lv_jumbo_workaround_ich8lan(hw, FALSE);
3280 	}
3281 
3282 	/* Make sure VLAN Filters are off */
3283 	rctl &= ~E1000_RCTL_VFE;
3284 
3285 	if (adapter->hw.mac.type < igb_mac_min) {
3286 		if (adapter->rx_mbuf_sz == MCLBYTES)
3287 			rctl |= E1000_RCTL_SZ_2048;
3288 		else if (adapter->rx_mbuf_sz == MJUMPAGESIZE)
3289 			rctl |= E1000_RCTL_SZ_4096 | E1000_RCTL_BSEX;
3290 		else if (adapter->rx_mbuf_sz > MJUMPAGESIZE)
3291 			rctl |= E1000_RCTL_SZ_8192 | E1000_RCTL_BSEX;
3292 
3293 		/* ensure we clear use DTYPE of 00 here */
3294 		rctl &= ~0x00000C00;
3295 	}
3296 
3297 	/* Write out the settings */
3298 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
3299 
3300 	return;
3301 }
3302 
3303 static void
3304 em_if_vlan_register(if_ctx_t ctx, u16 vtag)
3305 {
3306 	struct adapter *adapter = iflib_get_softc(ctx);
3307 	u32 index, bit;
3308 
3309 	index = (vtag >> 5) & 0x7F;
3310 	bit = vtag & 0x1F;
3311 	adapter->shadow_vfta[index] |= (1 << bit);
3312 	++adapter->num_vlans;
3313 }
3314 
3315 static void
3316 em_if_vlan_unregister(if_ctx_t ctx, u16 vtag)
3317 {
3318 	struct adapter *adapter = iflib_get_softc(ctx);
3319 	u32 index, bit;
3320 
3321 	index = (vtag >> 5) & 0x7F;
3322 	bit = vtag & 0x1F;
3323 	adapter->shadow_vfta[index] &= ~(1 << bit);
3324 	--adapter->num_vlans;
3325 }
3326 
3327 static void
3328 em_setup_vlan_hw_support(struct adapter *adapter)
3329 {
3330 	struct e1000_hw *hw = &adapter->hw;
3331 	u32 reg;
3332 
3333 	/*
3334 	 * We get here thru init_locked, meaning
3335 	 * a soft reset, this has already cleared
3336 	 * the VFTA and other state, so if there
3337 	 * have been no vlan's registered do nothing.
3338 	 */
3339 	if (adapter->num_vlans == 0)
3340 		return;
3341 
3342 	/*
3343 	 * A soft reset zero's out the VFTA, so
3344 	 * we need to repopulate it now.
3345 	 */
3346 	for (int i = 0; i < EM_VFTA_SIZE; i++)
3347 		if (adapter->shadow_vfta[i] != 0)
3348 			E1000_WRITE_REG_ARRAY(hw, E1000_VFTA,
3349 			    i, adapter->shadow_vfta[i]);
3350 
3351 	reg = E1000_READ_REG(hw, E1000_CTRL);
3352 	reg |= E1000_CTRL_VME;
3353 	E1000_WRITE_REG(hw, E1000_CTRL, reg);
3354 
3355 	/* Enable the Filter Table */
3356 	reg = E1000_READ_REG(hw, E1000_RCTL);
3357 	reg &= ~E1000_RCTL_CFIEN;
3358 	reg |= E1000_RCTL_VFE;
3359 	E1000_WRITE_REG(hw, E1000_RCTL, reg);
3360 }
3361 
3362 static void
3363 em_if_enable_intr(if_ctx_t ctx)
3364 {
3365 	struct adapter *adapter = iflib_get_softc(ctx);
3366 	struct e1000_hw *hw = &adapter->hw;
3367 	u32 ims_mask = IMS_ENABLE_MASK;
3368 
3369 	if (hw->mac.type == e1000_82574) {
3370 		E1000_WRITE_REG(hw, EM_EIAC, EM_MSIX_MASK);
3371 		ims_mask |= adapter->ims;
3372 	} else if (adapter->intr_type == IFLIB_INTR_MSIX && hw->mac.type >= igb_mac_min)  {
3373 		u32 mask = (adapter->que_mask | adapter->link_mask);
3374 
3375 		E1000_WRITE_REG(&adapter->hw, E1000_EIAC, mask);
3376 		E1000_WRITE_REG(&adapter->hw, E1000_EIAM, mask);
3377 		E1000_WRITE_REG(&adapter->hw, E1000_EIMS, mask);
3378 		ims_mask = E1000_IMS_LSC;
3379 	}
3380 
3381 	E1000_WRITE_REG(hw, E1000_IMS, ims_mask);
3382 }
3383 
3384 static void
3385 em_if_disable_intr(if_ctx_t ctx)
3386 {
3387 	struct adapter *adapter = iflib_get_softc(ctx);
3388 	struct e1000_hw *hw = &adapter->hw;
3389 
3390 	if (adapter->intr_type == IFLIB_INTR_MSIX) {
3391 		if (hw->mac.type >= igb_mac_min)
3392 			E1000_WRITE_REG(&adapter->hw, E1000_EIMC, ~0);
3393 		E1000_WRITE_REG(&adapter->hw, E1000_EIAC, 0);
3394 	}
3395 	E1000_WRITE_REG(&adapter->hw, E1000_IMC, 0xffffffff);
3396 }
3397 
3398 /*
3399  * Bit of a misnomer, what this really means is
3400  * to enable OS management of the system... aka
3401  * to disable special hardware management features
3402  */
3403 static void
3404 em_init_manageability(struct adapter *adapter)
3405 {
3406 	/* A shared code workaround */
3407 #define E1000_82542_MANC2H E1000_MANC2H
3408 	if (adapter->has_manage) {
3409 		int manc2h = E1000_READ_REG(&adapter->hw, E1000_MANC2H);
3410 		int manc = E1000_READ_REG(&adapter->hw, E1000_MANC);
3411 
3412 		/* disable hardware interception of ARP */
3413 		manc &= ~(E1000_MANC_ARP_EN);
3414 
3415 		/* enable receiving management packets to the host */
3416 		manc |= E1000_MANC_EN_MNG2HOST;
3417 #define E1000_MNG2HOST_PORT_623 (1 << 5)
3418 #define E1000_MNG2HOST_PORT_664 (1 << 6)
3419 		manc2h |= E1000_MNG2HOST_PORT_623;
3420 		manc2h |= E1000_MNG2HOST_PORT_664;
3421 		E1000_WRITE_REG(&adapter->hw, E1000_MANC2H, manc2h);
3422 		E1000_WRITE_REG(&adapter->hw, E1000_MANC, manc);
3423 	}
3424 }
3425 
3426 /*
3427  * Give control back to hardware management
3428  * controller if there is one.
3429  */
3430 static void
3431 em_release_manageability(struct adapter *adapter)
3432 {
3433 	if (adapter->has_manage) {
3434 		int manc = E1000_READ_REG(&adapter->hw, E1000_MANC);
3435 
3436 		/* re-enable hardware interception of ARP */
3437 		manc |= E1000_MANC_ARP_EN;
3438 		manc &= ~E1000_MANC_EN_MNG2HOST;
3439 
3440 		E1000_WRITE_REG(&adapter->hw, E1000_MANC, manc);
3441 	}
3442 }
3443 
3444 /*
3445  * em_get_hw_control sets the {CTRL_EXT|FWSM}:DRV_LOAD bit.
3446  * For ASF and Pass Through versions of f/w this means
3447  * that the driver is loaded. For AMT version type f/w
3448  * this means that the network i/f is open.
3449  */
3450 static void
3451 em_get_hw_control(struct adapter *adapter)
3452 {
3453 	u32 ctrl_ext, swsm;
3454 
3455 	if (adapter->vf_ifp)
3456 		return;
3457 
3458 	if (adapter->hw.mac.type == e1000_82573) {
3459 		swsm = E1000_READ_REG(&adapter->hw, E1000_SWSM);
3460 		E1000_WRITE_REG(&adapter->hw, E1000_SWSM,
3461 		    swsm | E1000_SWSM_DRV_LOAD);
3462 		return;
3463 	}
3464 	/* else */
3465 	ctrl_ext = E1000_READ_REG(&adapter->hw, E1000_CTRL_EXT);
3466 	E1000_WRITE_REG(&adapter->hw, E1000_CTRL_EXT,
3467 	    ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
3468 }
3469 
3470 /*
3471  * em_release_hw_control resets {CTRL_EXT|FWSM}:DRV_LOAD bit.
3472  * For ASF and Pass Through versions of f/w this means that
3473  * the driver is no longer loaded. For AMT versions of the
3474  * f/w this means that the network i/f is closed.
3475  */
3476 static void
3477 em_release_hw_control(struct adapter *adapter)
3478 {
3479 	u32 ctrl_ext, swsm;
3480 
3481 	if (!adapter->has_manage)
3482 		return;
3483 
3484 	if (adapter->hw.mac.type == e1000_82573) {
3485 		swsm = E1000_READ_REG(&adapter->hw, E1000_SWSM);
3486 		E1000_WRITE_REG(&adapter->hw, E1000_SWSM,
3487 		    swsm & ~E1000_SWSM_DRV_LOAD);
3488 		return;
3489 	}
3490 	/* else */
3491 	ctrl_ext = E1000_READ_REG(&adapter->hw, E1000_CTRL_EXT);
3492 	E1000_WRITE_REG(&adapter->hw, E1000_CTRL_EXT,
3493 	    ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
3494 	return;
3495 }
3496 
3497 static int
3498 em_is_valid_ether_addr(u8 *addr)
3499 {
3500 	char zero_addr[6] = { 0, 0, 0, 0, 0, 0 };
3501 
3502 	if ((addr[0] & 1) || (!bcmp(addr, zero_addr, ETHER_ADDR_LEN))) {
3503 		return (FALSE);
3504 	}
3505 
3506 	return (TRUE);
3507 }
3508 
3509 /*
3510 ** Parse the interface capabilities with regard
3511 ** to both system management and wake-on-lan for
3512 ** later use.
3513 */
3514 static void
3515 em_get_wakeup(if_ctx_t ctx)
3516 {
3517 	struct adapter *adapter = iflib_get_softc(ctx);
3518 	device_t dev = iflib_get_dev(ctx);
3519 	u16 eeprom_data = 0, device_id, apme_mask;
3520 
3521 	adapter->has_manage = e1000_enable_mng_pass_thru(&adapter->hw);
3522 	apme_mask = EM_EEPROM_APME;
3523 
3524 	switch (adapter->hw.mac.type) {
3525 	case e1000_82542:
3526 	case e1000_82543:
3527 		break;
3528 	case e1000_82544:
3529 		e1000_read_nvm(&adapter->hw,
3530 		    NVM_INIT_CONTROL2_REG, 1, &eeprom_data);
3531 		apme_mask = EM_82544_APME;
3532 		break;
3533 	case e1000_82546:
3534 	case e1000_82546_rev_3:
3535 		if (adapter->hw.bus.func == 1) {
3536 			e1000_read_nvm(&adapter->hw,
3537 			    NVM_INIT_CONTROL3_PORT_B, 1, &eeprom_data);
3538 			break;
3539 		} else
3540 			e1000_read_nvm(&adapter->hw,
3541 			    NVM_INIT_CONTROL3_PORT_A, 1, &eeprom_data);
3542 		break;
3543 	case e1000_82573:
3544 	case e1000_82583:
3545 		adapter->has_amt = TRUE;
3546 		/* FALLTHROUGH */
3547 	case e1000_82571:
3548 	case e1000_82572:
3549 	case e1000_80003es2lan:
3550 		if (adapter->hw.bus.func == 1) {
3551 			e1000_read_nvm(&adapter->hw,
3552 			    NVM_INIT_CONTROL3_PORT_B, 1, &eeprom_data);
3553 			break;
3554 		} else
3555 			e1000_read_nvm(&adapter->hw,
3556 			    NVM_INIT_CONTROL3_PORT_A, 1, &eeprom_data);
3557 		break;
3558 	case e1000_ich8lan:
3559 	case e1000_ich9lan:
3560 	case e1000_ich10lan:
3561 	case e1000_pchlan:
3562 	case e1000_pch2lan:
3563 	case e1000_pch_lpt:
3564 	case e1000_pch_spt:
3565 	case e1000_82575:	/* listing all igb devices */
3566 	case e1000_82576:
3567 	case e1000_82580:
3568 	case e1000_i350:
3569 	case e1000_i354:
3570 	case e1000_i210:
3571 	case e1000_i211:
3572 	case e1000_vfadapt:
3573 	case e1000_vfadapt_i350:
3574 		apme_mask = E1000_WUC_APME;
3575 		adapter->has_amt = TRUE;
3576 		eeprom_data = E1000_READ_REG(&adapter->hw, E1000_WUC);
3577 		break;
3578 	default:
3579 		e1000_read_nvm(&adapter->hw,
3580 		    NVM_INIT_CONTROL3_PORT_A, 1, &eeprom_data);
3581 		break;
3582 	}
3583 	if (eeprom_data & apme_mask)
3584 		adapter->wol = (E1000_WUFC_MAG | E1000_WUFC_MC);
3585 	/*
3586 	 * We have the eeprom settings, now apply the special cases
3587 	 * where the eeprom may be wrong or the board won't support
3588 	 * wake on lan on a particular port
3589 	 */
3590 	device_id = pci_get_device(dev);
3591 	switch (device_id) {
3592 	case E1000_DEV_ID_82546GB_PCIE:
3593 		adapter->wol = 0;
3594 		break;
3595 	case E1000_DEV_ID_82546EB_FIBER:
3596 	case E1000_DEV_ID_82546GB_FIBER:
3597 		/* Wake events only supported on port A for dual fiber
3598 		 * regardless of eeprom setting */
3599 		if (E1000_READ_REG(&adapter->hw, E1000_STATUS) &
3600 		    E1000_STATUS_FUNC_1)
3601 			adapter->wol = 0;
3602 		break;
3603 	case E1000_DEV_ID_82546GB_QUAD_COPPER_KSP3:
3604 		/* if quad port adapter, disable WoL on all but port A */
3605 		if (global_quad_port_a != 0)
3606 			adapter->wol = 0;
3607 		/* Reset for multiple quad port adapters */
3608 		if (++global_quad_port_a == 4)
3609 			global_quad_port_a = 0;
3610 		break;
3611 	case E1000_DEV_ID_82571EB_FIBER:
3612 		/* Wake events only supported on port A for dual fiber
3613 		 * regardless of eeprom setting */
3614 		if (E1000_READ_REG(&adapter->hw, E1000_STATUS) &
3615 		    E1000_STATUS_FUNC_1)
3616 			adapter->wol = 0;
3617 		break;
3618 	case E1000_DEV_ID_82571EB_QUAD_COPPER:
3619 	case E1000_DEV_ID_82571EB_QUAD_FIBER:
3620 	case E1000_DEV_ID_82571EB_QUAD_COPPER_LP:
3621 		/* if quad port adapter, disable WoL on all but port A */
3622 		if (global_quad_port_a != 0)
3623 			adapter->wol = 0;
3624 		/* Reset for multiple quad port adapters */
3625 		if (++global_quad_port_a == 4)
3626 			global_quad_port_a = 0;
3627 		break;
3628 	}
3629 	return;
3630 }
3631 
3632 
3633 /*
3634  * Enable PCI Wake On Lan capability
3635  */
3636 static void
3637 em_enable_wakeup(if_ctx_t ctx)
3638 {
3639 	struct adapter *adapter = iflib_get_softc(ctx);
3640 	device_t dev = iflib_get_dev(ctx);
3641 	if_t ifp = iflib_get_ifp(ctx);
3642 	u32 pmc, ctrl, ctrl_ext, rctl, wuc;
3643 	u16 status;
3644 
3645 	if ((pci_find_cap(dev, PCIY_PMG, &pmc) != 0))
3646 		return;
3647 
3648 	/* Advertise the wakeup capability */
3649 	ctrl = E1000_READ_REG(&adapter->hw, E1000_CTRL);
3650 	ctrl |= (E1000_CTRL_SWDPIN2 | E1000_CTRL_SWDPIN3);
3651 	E1000_WRITE_REG(&adapter->hw, E1000_CTRL, ctrl);
3652 	wuc = E1000_READ_REG(&adapter->hw, E1000_WUC);
3653 	wuc |= (E1000_WUC_PME_EN | E1000_WUC_APME);
3654 	E1000_WRITE_REG(&adapter->hw, E1000_WUC, wuc);
3655 
3656 	if ((adapter->hw.mac.type == e1000_ich8lan) ||
3657 	    (adapter->hw.mac.type == e1000_pchlan) ||
3658 	    (adapter->hw.mac.type == e1000_ich9lan) ||
3659 	    (adapter->hw.mac.type == e1000_ich10lan))
3660 		e1000_suspend_workarounds_ich8lan(&adapter->hw);
3661 
3662 	/* Keep the laser running on Fiber adapters */
3663 	if (adapter->hw.phy.media_type == e1000_media_type_fiber ||
3664 	    adapter->hw.phy.media_type == e1000_media_type_internal_serdes) {
3665 		ctrl_ext = E1000_READ_REG(&adapter->hw, E1000_CTRL_EXT);
3666 		ctrl_ext |= E1000_CTRL_EXT_SDP3_DATA;
3667 		E1000_WRITE_REG(&adapter->hw, E1000_CTRL_EXT, ctrl_ext);
3668 	}
3669 
3670 	/*
3671 	 * Determine type of Wakeup: note that wol
3672 	 * is set with all bits on by default.
3673 	 */
3674 	if ((if_getcapenable(ifp) & IFCAP_WOL_MAGIC) == 0)
3675 		adapter->wol &= ~E1000_WUFC_MAG;
3676 
3677 	if ((if_getcapenable(ifp) & IFCAP_WOL_UCAST) == 0)
3678 		adapter->wol &= ~E1000_WUFC_EX;
3679 
3680 	if ((if_getcapenable(ifp) & IFCAP_WOL_MCAST) == 0)
3681 		adapter->wol &= ~E1000_WUFC_MC;
3682 	else {
3683 		rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
3684 		rctl |= E1000_RCTL_MPE;
3685 		E1000_WRITE_REG(&adapter->hw, E1000_RCTL, rctl);
3686 	}
3687 
3688 	if ( adapter->hw.mac.type >= e1000_pchlan) {
3689 		if (em_enable_phy_wakeup(adapter))
3690 			return;
3691 	} else {
3692 		E1000_WRITE_REG(&adapter->hw, E1000_WUC, E1000_WUC_PME_EN);
3693 		E1000_WRITE_REG(&adapter->hw, E1000_WUFC, adapter->wol);
3694 	}
3695 
3696 	if (adapter->hw.phy.type == e1000_phy_igp_3)
3697 		e1000_igp3_phy_powerdown_workaround_ich8lan(&adapter->hw);
3698 
3699 	/* Request PME */
3700 	status = pci_read_config(dev, pmc + PCIR_POWER_STATUS, 2);
3701 	status &= ~(PCIM_PSTAT_PME | PCIM_PSTAT_PMEENABLE);
3702 	if (if_getcapenable(ifp) & IFCAP_WOL)
3703 		status |= PCIM_PSTAT_PME | PCIM_PSTAT_PMEENABLE;
3704 	pci_write_config(dev, pmc + PCIR_POWER_STATUS, status, 2);
3705 
3706 	return;
3707 }
3708 
3709 /*
3710  * WOL in the newer chipset interfaces (pchlan)
3711  * require thing to be copied into the phy
3712  */
3713 static int
3714 em_enable_phy_wakeup(struct adapter *adapter)
3715 {
3716 	struct e1000_hw *hw = &adapter->hw;
3717 	u32 mreg, ret = 0;
3718 	u16 preg;
3719 
3720 	/* copy MAC RARs to PHY RARs */
3721 	e1000_copy_rx_addrs_to_phy_ich8lan(hw);
3722 
3723 	/* copy MAC MTA to PHY MTA */
3724 	for (int i = 0; i < adapter->hw.mac.mta_reg_count; i++) {
3725 		mreg = E1000_READ_REG_ARRAY(hw, E1000_MTA, i);
3726 		e1000_write_phy_reg(hw, BM_MTA(i), (u16)(mreg & 0xFFFF));
3727 		e1000_write_phy_reg(hw, BM_MTA(i) + 1,
3728 		    (u16)((mreg >> 16) & 0xFFFF));
3729 	}
3730 
3731 	/* configure PHY Rx Control register */
3732 	e1000_read_phy_reg(&adapter->hw, BM_RCTL, &preg);
3733 	mreg = E1000_READ_REG(hw, E1000_RCTL);
3734 	if (mreg & E1000_RCTL_UPE)
3735 		preg |= BM_RCTL_UPE;
3736 	if (mreg & E1000_RCTL_MPE)
3737 		preg |= BM_RCTL_MPE;
3738 	preg &= ~(BM_RCTL_MO_MASK);
3739 	if (mreg & E1000_RCTL_MO_3)
3740 		preg |= (((mreg & E1000_RCTL_MO_3) >> E1000_RCTL_MO_SHIFT)
3741 				<< BM_RCTL_MO_SHIFT);
3742 	if (mreg & E1000_RCTL_BAM)
3743 		preg |= BM_RCTL_BAM;
3744 	if (mreg & E1000_RCTL_PMCF)
3745 		preg |= BM_RCTL_PMCF;
3746 	mreg = E1000_READ_REG(hw, E1000_CTRL);
3747 	if (mreg & E1000_CTRL_RFCE)
3748 		preg |= BM_RCTL_RFCE;
3749 	e1000_write_phy_reg(&adapter->hw, BM_RCTL, preg);
3750 
3751 	/* enable PHY wakeup in MAC register */
3752 	E1000_WRITE_REG(hw, E1000_WUC,
3753 	    E1000_WUC_PHY_WAKE | E1000_WUC_PME_EN | E1000_WUC_APME);
3754 	E1000_WRITE_REG(hw, E1000_WUFC, adapter->wol);
3755 
3756 	/* configure and enable PHY wakeup in PHY registers */
3757 	e1000_write_phy_reg(&adapter->hw, BM_WUFC, adapter->wol);
3758 	e1000_write_phy_reg(&adapter->hw, BM_WUC, E1000_WUC_PME_EN);
3759 
3760 	/* activate PHY wakeup */
3761 	ret = hw->phy.ops.acquire(hw);
3762 	if (ret) {
3763 		printf("Could not acquire PHY\n");
3764 		return ret;
3765 	}
3766 	e1000_write_phy_reg_mdic(hw, IGP01E1000_PHY_PAGE_SELECT,
3767 	                         (BM_WUC_ENABLE_PAGE << IGP_PAGE_SHIFT));
3768 	ret = e1000_read_phy_reg_mdic(hw, BM_WUC_ENABLE_REG, &preg);
3769 	if (ret) {
3770 		printf("Could not read PHY page 769\n");
3771 		goto out;
3772 	}
3773 	preg |= BM_WUC_ENABLE_BIT | BM_WUC_HOST_WU_BIT;
3774 	ret = e1000_write_phy_reg_mdic(hw, BM_WUC_ENABLE_REG, preg);
3775 	if (ret)
3776 		printf("Could not set PHY Host Wakeup bit\n");
3777 out:
3778 	hw->phy.ops.release(hw);
3779 
3780 	return ret;
3781 }
3782 
3783 static void
3784 em_if_led_func(if_ctx_t ctx, int onoff)
3785 {
3786 	struct adapter *adapter = iflib_get_softc(ctx);
3787 
3788 	if (onoff) {
3789 		e1000_setup_led(&adapter->hw);
3790 		e1000_led_on(&adapter->hw);
3791 	} else {
3792 		e1000_led_off(&adapter->hw);
3793 		e1000_cleanup_led(&adapter->hw);
3794 	}
3795 }
3796 
3797 /*
3798  * Disable the L0S and L1 LINK states
3799  */
3800 static void
3801 em_disable_aspm(struct adapter *adapter)
3802 {
3803 	int base, reg;
3804 	u16 link_cap,link_ctrl;
3805 	device_t dev = adapter->dev;
3806 
3807 	switch (adapter->hw.mac.type) {
3808 	case e1000_82573:
3809 	case e1000_82574:
3810 	case e1000_82583:
3811 		break;
3812 	default:
3813 		return;
3814 	}
3815 	if (pci_find_cap(dev, PCIY_EXPRESS, &base) != 0)
3816 		return;
3817 	reg = base + PCIER_LINK_CAP;
3818 	link_cap = pci_read_config(dev, reg, 2);
3819 	if ((link_cap & PCIEM_LINK_CAP_ASPM) == 0)
3820 		return;
3821 	reg = base + PCIER_LINK_CTL;
3822 	link_ctrl = pci_read_config(dev, reg, 2);
3823 	link_ctrl &= ~PCIEM_LINK_CTL_ASPMC;
3824 	pci_write_config(dev, reg, link_ctrl, 2);
3825 	return;
3826 }
3827 
3828 /**********************************************************************
3829  *
3830  *  Update the board statistics counters.
3831  *
3832  **********************************************************************/
3833 static void
3834 em_update_stats_counters(struct adapter *adapter)
3835 {
3836 
3837 	if(adapter->hw.phy.media_type == e1000_media_type_copper ||
3838 	   (E1000_READ_REG(&adapter->hw, E1000_STATUS) & E1000_STATUS_LU)) {
3839 		adapter->stats.symerrs += E1000_READ_REG(&adapter->hw, E1000_SYMERRS);
3840 		adapter->stats.sec += E1000_READ_REG(&adapter->hw, E1000_SEC);
3841 	}
3842 	adapter->stats.crcerrs += E1000_READ_REG(&adapter->hw, E1000_CRCERRS);
3843 	adapter->stats.mpc += E1000_READ_REG(&adapter->hw, E1000_MPC);
3844 	adapter->stats.scc += E1000_READ_REG(&adapter->hw, E1000_SCC);
3845 	adapter->stats.ecol += E1000_READ_REG(&adapter->hw, E1000_ECOL);
3846 
3847 	adapter->stats.mcc += E1000_READ_REG(&adapter->hw, E1000_MCC);
3848 	adapter->stats.latecol += E1000_READ_REG(&adapter->hw, E1000_LATECOL);
3849 	adapter->stats.colc += E1000_READ_REG(&adapter->hw, E1000_COLC);
3850 	adapter->stats.dc += E1000_READ_REG(&adapter->hw, E1000_DC);
3851 	adapter->stats.rlec += E1000_READ_REG(&adapter->hw, E1000_RLEC);
3852 	adapter->stats.xonrxc += E1000_READ_REG(&adapter->hw, E1000_XONRXC);
3853 	adapter->stats.xontxc += E1000_READ_REG(&adapter->hw, E1000_XONTXC);
3854 	adapter->stats.xoffrxc += E1000_READ_REG(&adapter->hw, E1000_XOFFRXC);
3855 	/*
3856 	 ** For watchdog management we need to know if we have been
3857 	 ** paused during the last interval, so capture that here.
3858 	*/
3859 	adapter->shared->isc_pause_frames = adapter->stats.xoffrxc;
3860 	adapter->stats.xofftxc += E1000_READ_REG(&adapter->hw, E1000_XOFFTXC);
3861 	adapter->stats.fcruc += E1000_READ_REG(&adapter->hw, E1000_FCRUC);
3862 	adapter->stats.prc64 += E1000_READ_REG(&adapter->hw, E1000_PRC64);
3863 	adapter->stats.prc127 += E1000_READ_REG(&adapter->hw, E1000_PRC127);
3864 	adapter->stats.prc255 += E1000_READ_REG(&adapter->hw, E1000_PRC255);
3865 	adapter->stats.prc511 += E1000_READ_REG(&adapter->hw, E1000_PRC511);
3866 	adapter->stats.prc1023 += E1000_READ_REG(&adapter->hw, E1000_PRC1023);
3867 	adapter->stats.prc1522 += E1000_READ_REG(&adapter->hw, E1000_PRC1522);
3868 	adapter->stats.gprc += E1000_READ_REG(&adapter->hw, E1000_GPRC);
3869 	adapter->stats.bprc += E1000_READ_REG(&adapter->hw, E1000_BPRC);
3870 	adapter->stats.mprc += E1000_READ_REG(&adapter->hw, E1000_MPRC);
3871 	adapter->stats.gptc += E1000_READ_REG(&adapter->hw, E1000_GPTC);
3872 
3873 	/* For the 64-bit byte counters the low dword must be read first. */
3874 	/* Both registers clear on the read of the high dword */
3875 
3876 	adapter->stats.gorc += E1000_READ_REG(&adapter->hw, E1000_GORCL) +
3877 	    ((u64)E1000_READ_REG(&adapter->hw, E1000_GORCH) << 32);
3878 	adapter->stats.gotc += E1000_READ_REG(&adapter->hw, E1000_GOTCL) +
3879 	    ((u64)E1000_READ_REG(&adapter->hw, E1000_GOTCH) << 32);
3880 
3881 	adapter->stats.rnbc += E1000_READ_REG(&adapter->hw, E1000_RNBC);
3882 	adapter->stats.ruc += E1000_READ_REG(&adapter->hw, E1000_RUC);
3883 	adapter->stats.rfc += E1000_READ_REG(&adapter->hw, E1000_RFC);
3884 	adapter->stats.roc += E1000_READ_REG(&adapter->hw, E1000_ROC);
3885 	adapter->stats.rjc += E1000_READ_REG(&adapter->hw, E1000_RJC);
3886 
3887 	adapter->stats.tor += E1000_READ_REG(&adapter->hw, E1000_TORH);
3888 	adapter->stats.tot += E1000_READ_REG(&adapter->hw, E1000_TOTH);
3889 
3890 	adapter->stats.tpr += E1000_READ_REG(&adapter->hw, E1000_TPR);
3891 	adapter->stats.tpt += E1000_READ_REG(&adapter->hw, E1000_TPT);
3892 	adapter->stats.ptc64 += E1000_READ_REG(&adapter->hw, E1000_PTC64);
3893 	adapter->stats.ptc127 += E1000_READ_REG(&adapter->hw, E1000_PTC127);
3894 	adapter->stats.ptc255 += E1000_READ_REG(&adapter->hw, E1000_PTC255);
3895 	adapter->stats.ptc511 += E1000_READ_REG(&adapter->hw, E1000_PTC511);
3896 	adapter->stats.ptc1023 += E1000_READ_REG(&adapter->hw, E1000_PTC1023);
3897 	adapter->stats.ptc1522 += E1000_READ_REG(&adapter->hw, E1000_PTC1522);
3898 	adapter->stats.mptc += E1000_READ_REG(&adapter->hw, E1000_MPTC);
3899 	adapter->stats.bptc += E1000_READ_REG(&adapter->hw, E1000_BPTC);
3900 
3901 	/* Interrupt Counts */
3902 
3903 	adapter->stats.iac += E1000_READ_REG(&adapter->hw, E1000_IAC);
3904 	adapter->stats.icrxptc += E1000_READ_REG(&adapter->hw, E1000_ICRXPTC);
3905 	adapter->stats.icrxatc += E1000_READ_REG(&adapter->hw, E1000_ICRXATC);
3906 	adapter->stats.ictxptc += E1000_READ_REG(&adapter->hw, E1000_ICTXPTC);
3907 	adapter->stats.ictxatc += E1000_READ_REG(&adapter->hw, E1000_ICTXATC);
3908 	adapter->stats.ictxqec += E1000_READ_REG(&adapter->hw, E1000_ICTXQEC);
3909 	adapter->stats.ictxqmtc += E1000_READ_REG(&adapter->hw, E1000_ICTXQMTC);
3910 	adapter->stats.icrxdmtc += E1000_READ_REG(&adapter->hw, E1000_ICRXDMTC);
3911 	adapter->stats.icrxoc += E1000_READ_REG(&adapter->hw, E1000_ICRXOC);
3912 
3913 	if (adapter->hw.mac.type >= e1000_82543) {
3914 		adapter->stats.algnerrc +=
3915 		E1000_READ_REG(&adapter->hw, E1000_ALGNERRC);
3916 		adapter->stats.rxerrc +=
3917 		E1000_READ_REG(&adapter->hw, E1000_RXERRC);
3918 		adapter->stats.tncrs +=
3919 		E1000_READ_REG(&adapter->hw, E1000_TNCRS);
3920 		adapter->stats.cexterr +=
3921 		E1000_READ_REG(&adapter->hw, E1000_CEXTERR);
3922 		adapter->stats.tsctc +=
3923 		E1000_READ_REG(&adapter->hw, E1000_TSCTC);
3924 		adapter->stats.tsctfc +=
3925 		E1000_READ_REG(&adapter->hw, E1000_TSCTFC);
3926 	}
3927 }
3928 
3929 static uint64_t
3930 em_if_get_counter(if_ctx_t ctx, ift_counter cnt)
3931 {
3932 	struct adapter *adapter = iflib_get_softc(ctx);
3933 	struct ifnet *ifp = iflib_get_ifp(ctx);
3934 
3935 	switch (cnt) {
3936 	case IFCOUNTER_COLLISIONS:
3937 		return (adapter->stats.colc);
3938 	case IFCOUNTER_IERRORS:
3939 		return (adapter->dropped_pkts + adapter->stats.rxerrc +
3940 		    adapter->stats.crcerrs + adapter->stats.algnerrc +
3941 		    adapter->stats.ruc + adapter->stats.roc +
3942 		    adapter->stats.mpc + adapter->stats.cexterr);
3943 	case IFCOUNTER_OERRORS:
3944 		return (adapter->stats.ecol + adapter->stats.latecol +
3945 		    adapter->watchdog_events);
3946 	default:
3947 		return (if_get_counter_default(ifp, cnt));
3948 	}
3949 }
3950 
3951 /* Export a single 32-bit register via a read-only sysctl. */
3952 static int
3953 em_sysctl_reg_handler(SYSCTL_HANDLER_ARGS)
3954 {
3955 	struct adapter *adapter;
3956 	u_int val;
3957 
3958 	adapter = oidp->oid_arg1;
3959 	val = E1000_READ_REG(&adapter->hw, oidp->oid_arg2);
3960 	return (sysctl_handle_int(oidp, &val, 0, req));
3961 }
3962 
3963 /*
3964  * Add sysctl variables, one per statistic, to the system.
3965  */
3966 static void
3967 em_add_hw_stats(struct adapter *adapter)
3968 {
3969 	device_t dev = iflib_get_dev(adapter->ctx);
3970 	struct em_tx_queue *tx_que = adapter->tx_queues;
3971 	struct em_rx_queue *rx_que = adapter->rx_queues;
3972 
3973 	struct sysctl_ctx_list *ctx = device_get_sysctl_ctx(dev);
3974 	struct sysctl_oid *tree = device_get_sysctl_tree(dev);
3975 	struct sysctl_oid_list *child = SYSCTL_CHILDREN(tree);
3976 	struct e1000_hw_stats *stats = &adapter->stats;
3977 
3978 	struct sysctl_oid *stat_node, *queue_node, *int_node;
3979 	struct sysctl_oid_list *stat_list, *queue_list, *int_list;
3980 
3981 #define QUEUE_NAME_LEN 32
3982 	char namebuf[QUEUE_NAME_LEN];
3983 
3984 	/* Driver Statistics */
3985 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "dropped",
3986 			CTLFLAG_RD, &adapter->dropped_pkts,
3987 			"Driver dropped packets");
3988 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "link_irq",
3989 			CTLFLAG_RD, &adapter->link_irq,
3990 			"Link MSIX IRQ Handled");
3991 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "mbuf_defrag_fail",
3992 			 CTLFLAG_RD, &adapter->mbuf_defrag_failed,
3993 			 "Defragmenting mbuf chain failed");
3994 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "tx_dma_fail",
3995 			CTLFLAG_RD, &adapter->no_tx_dma_setup,
3996 			"Driver tx dma failure in xmit");
3997 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "rx_overruns",
3998 			CTLFLAG_RD, &adapter->rx_overruns,
3999 			"RX overruns");
4000 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "watchdog_timeouts",
4001 			CTLFLAG_RD, &adapter->watchdog_events,
4002 			"Watchdog timeouts");
4003 	SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "device_control",
4004 			CTLTYPE_UINT | CTLFLAG_RD, adapter, E1000_CTRL,
4005 			em_sysctl_reg_handler, "IU",
4006 			"Device Control Register");
4007 	SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "rx_control",
4008 			CTLTYPE_UINT | CTLFLAG_RD, adapter, E1000_RCTL,
4009 			em_sysctl_reg_handler, "IU",
4010 			"Receiver Control Register");
4011 	SYSCTL_ADD_UINT(ctx, child, OID_AUTO, "fc_high_water",
4012 			CTLFLAG_RD, &adapter->hw.fc.high_water, 0,
4013 			"Flow Control High Watermark");
4014 	SYSCTL_ADD_UINT(ctx, child, OID_AUTO, "fc_low_water",
4015 			CTLFLAG_RD, &adapter->hw.fc.low_water, 0,
4016 			"Flow Control Low Watermark");
4017 
4018 	for (int i = 0; i < adapter->tx_num_queues; i++, tx_que++) {
4019 		struct tx_ring *txr = &tx_que->txr;
4020 		snprintf(namebuf, QUEUE_NAME_LEN, "queue_tx_%d", i);
4021 		queue_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, namebuf,
4022 					    CTLFLAG_RD, NULL, "TX Queue Name");
4023 		queue_list = SYSCTL_CHILDREN(queue_node);
4024 
4025 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "txd_head",
4026 				CTLTYPE_UINT | CTLFLAG_RD, adapter,
4027 				E1000_TDH(txr->me),
4028 				em_sysctl_reg_handler, "IU",
4029 				"Transmit Descriptor Head");
4030 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "txd_tail",
4031 				CTLTYPE_UINT | CTLFLAG_RD, adapter,
4032 				E1000_TDT(txr->me),
4033 				em_sysctl_reg_handler, "IU",
4034 				"Transmit Descriptor Tail");
4035 		SYSCTL_ADD_ULONG(ctx, queue_list, OID_AUTO, "tx_irq",
4036 				CTLFLAG_RD, &txr->tx_irq,
4037 				"Queue MSI-X Transmit Interrupts");
4038 	}
4039 
4040 	for (int j = 0; j < adapter->rx_num_queues; j++, rx_que++) {
4041 		struct rx_ring *rxr = &rx_que->rxr;
4042 		snprintf(namebuf, QUEUE_NAME_LEN, "queue_rx_%d", j);
4043 		queue_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, namebuf,
4044 					    CTLFLAG_RD, NULL, "RX Queue Name");
4045 		queue_list = SYSCTL_CHILDREN(queue_node);
4046 
4047 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "rxd_head",
4048 				CTLTYPE_UINT | CTLFLAG_RD, adapter,
4049 				E1000_RDH(rxr->me),
4050 				em_sysctl_reg_handler, "IU",
4051 				"Receive Descriptor Head");
4052 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "rxd_tail",
4053 				CTLTYPE_UINT | CTLFLAG_RD, adapter,
4054 				E1000_RDT(rxr->me),
4055 				em_sysctl_reg_handler, "IU",
4056 				"Receive Descriptor Tail");
4057 		SYSCTL_ADD_ULONG(ctx, queue_list, OID_AUTO, "rx_irq",
4058 				CTLFLAG_RD, &rxr->rx_irq,
4059 				"Queue MSI-X Receive Interrupts");
4060 	}
4061 
4062 	/* MAC stats get their own sub node */
4063 
4064 	stat_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, "mac_stats",
4065 				    CTLFLAG_RD, NULL, "Statistics");
4066 	stat_list = SYSCTL_CHILDREN(stat_node);
4067 
4068 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "excess_coll",
4069 			CTLFLAG_RD, &stats->ecol,
4070 			"Excessive collisions");
4071 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "single_coll",
4072 			CTLFLAG_RD, &stats->scc,
4073 			"Single collisions");
4074 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "multiple_coll",
4075 			CTLFLAG_RD, &stats->mcc,
4076 			"Multiple collisions");
4077 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "late_coll",
4078 			CTLFLAG_RD, &stats->latecol,
4079 			"Late collisions");
4080 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "collision_count",
4081 			CTLFLAG_RD, &stats->colc,
4082 			"Collision Count");
4083 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "symbol_errors",
4084 			CTLFLAG_RD, &adapter->stats.symerrs,
4085 			"Symbol Errors");
4086 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "sequence_errors",
4087 			CTLFLAG_RD, &adapter->stats.sec,
4088 			"Sequence Errors");
4089 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "defer_count",
4090 			CTLFLAG_RD, &adapter->stats.dc,
4091 			"Defer Count");
4092 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "missed_packets",
4093 			CTLFLAG_RD, &adapter->stats.mpc,
4094 			"Missed Packets");
4095 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_no_buff",
4096 			CTLFLAG_RD, &adapter->stats.rnbc,
4097 			"Receive No Buffers");
4098 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_undersize",
4099 			CTLFLAG_RD, &adapter->stats.ruc,
4100 			"Receive Undersize");
4101 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_fragmented",
4102 			CTLFLAG_RD, &adapter->stats.rfc,
4103 			"Fragmented Packets Received ");
4104 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_oversize",
4105 			CTLFLAG_RD, &adapter->stats.roc,
4106 			"Oversized Packets Received");
4107 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_jabber",
4108 			CTLFLAG_RD, &adapter->stats.rjc,
4109 			"Recevied Jabber");
4110 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_errs",
4111 			CTLFLAG_RD, &adapter->stats.rxerrc,
4112 			"Receive Errors");
4113 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "crc_errs",
4114 			CTLFLAG_RD, &adapter->stats.crcerrs,
4115 			"CRC errors");
4116 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "alignment_errs",
4117 			CTLFLAG_RD, &adapter->stats.algnerrc,
4118 			"Alignment Errors");
4119 	/* On 82575 these are collision counts */
4120 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "coll_ext_errs",
4121 			CTLFLAG_RD, &adapter->stats.cexterr,
4122 			"Collision/Carrier extension errors");
4123 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xon_recvd",
4124 			CTLFLAG_RD, &adapter->stats.xonrxc,
4125 			"XON Received");
4126 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xon_txd",
4127 			CTLFLAG_RD, &adapter->stats.xontxc,
4128 			"XON Transmitted");
4129 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xoff_recvd",
4130 			CTLFLAG_RD, &adapter->stats.xoffrxc,
4131 			"XOFF Received");
4132 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xoff_txd",
4133 			CTLFLAG_RD, &adapter->stats.xofftxc,
4134 			"XOFF Transmitted");
4135 
4136 	/* Packet Reception Stats */
4137 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_pkts_recvd",
4138 			CTLFLAG_RD, &adapter->stats.tpr,
4139 			"Total Packets Received ");
4140 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_pkts_recvd",
4141 			CTLFLAG_RD, &adapter->stats.gprc,
4142 			"Good Packets Received");
4143 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "bcast_pkts_recvd",
4144 			CTLFLAG_RD, &adapter->stats.bprc,
4145 			"Broadcast Packets Received");
4146 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "mcast_pkts_recvd",
4147 			CTLFLAG_RD, &adapter->stats.mprc,
4148 			"Multicast Packets Received");
4149 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_64",
4150 			CTLFLAG_RD, &adapter->stats.prc64,
4151 			"64 byte frames received ");
4152 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_65_127",
4153 			CTLFLAG_RD, &adapter->stats.prc127,
4154 			"65-127 byte frames received");
4155 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_128_255",
4156 			CTLFLAG_RD, &adapter->stats.prc255,
4157 			"128-255 byte frames received");
4158 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_256_511",
4159 			CTLFLAG_RD, &adapter->stats.prc511,
4160 			"256-511 byte frames received");
4161 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_512_1023",
4162 			CTLFLAG_RD, &adapter->stats.prc1023,
4163 			"512-1023 byte frames received");
4164 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_1024_1522",
4165 			CTLFLAG_RD, &adapter->stats.prc1522,
4166 			"1023-1522 byte frames received");
4167 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_octets_recvd",
4168 			CTLFLAG_RD, &adapter->stats.gorc,
4169 			"Good Octets Received");
4170 
4171 	/* Packet Transmission Stats */
4172 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_octets_txd",
4173 			CTLFLAG_RD, &adapter->stats.gotc,
4174 			"Good Octets Transmitted");
4175 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_pkts_txd",
4176 			CTLFLAG_RD, &adapter->stats.tpt,
4177 			"Total Packets Transmitted");
4178 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_pkts_txd",
4179 			CTLFLAG_RD, &adapter->stats.gptc,
4180 			"Good Packets Transmitted");
4181 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "bcast_pkts_txd",
4182 			CTLFLAG_RD, &adapter->stats.bptc,
4183 			"Broadcast Packets Transmitted");
4184 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "mcast_pkts_txd",
4185 			CTLFLAG_RD, &adapter->stats.mptc,
4186 			"Multicast Packets Transmitted");
4187 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_64",
4188 			CTLFLAG_RD, &adapter->stats.ptc64,
4189 			"64 byte frames transmitted ");
4190 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_65_127",
4191 			CTLFLAG_RD, &adapter->stats.ptc127,
4192 			"65-127 byte frames transmitted");
4193 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_128_255",
4194 			CTLFLAG_RD, &adapter->stats.ptc255,
4195 			"128-255 byte frames transmitted");
4196 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_256_511",
4197 			CTLFLAG_RD, &adapter->stats.ptc511,
4198 			"256-511 byte frames transmitted");
4199 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_512_1023",
4200 			CTLFLAG_RD, &adapter->stats.ptc1023,
4201 			"512-1023 byte frames transmitted");
4202 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_1024_1522",
4203 			CTLFLAG_RD, &adapter->stats.ptc1522,
4204 			"1024-1522 byte frames transmitted");
4205 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tso_txd",
4206 			CTLFLAG_RD, &adapter->stats.tsctc,
4207 			"TSO Contexts Transmitted");
4208 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tso_ctx_fail",
4209 			CTLFLAG_RD, &adapter->stats.tsctfc,
4210 			"TSO Contexts Failed");
4211 
4212 
4213 	/* Interrupt Stats */
4214 
4215 	int_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, "interrupts",
4216 				    CTLFLAG_RD, NULL, "Interrupt Statistics");
4217 	int_list = SYSCTL_CHILDREN(int_node);
4218 
4219 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "asserts",
4220 			CTLFLAG_RD, &adapter->stats.iac,
4221 			"Interrupt Assertion Count");
4222 
4223 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_pkt_timer",
4224 			CTLFLAG_RD, &adapter->stats.icrxptc,
4225 			"Interrupt Cause Rx Pkt Timer Expire Count");
4226 
4227 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_abs_timer",
4228 			CTLFLAG_RD, &adapter->stats.icrxatc,
4229 			"Interrupt Cause Rx Abs Timer Expire Count");
4230 
4231 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_pkt_timer",
4232 			CTLFLAG_RD, &adapter->stats.ictxptc,
4233 			"Interrupt Cause Tx Pkt Timer Expire Count");
4234 
4235 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_abs_timer",
4236 			CTLFLAG_RD, &adapter->stats.ictxatc,
4237 			"Interrupt Cause Tx Abs Timer Expire Count");
4238 
4239 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_queue_empty",
4240 			CTLFLAG_RD, &adapter->stats.ictxqec,
4241 			"Interrupt Cause Tx Queue Empty Count");
4242 
4243 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_queue_min_thresh",
4244 			CTLFLAG_RD, &adapter->stats.ictxqmtc,
4245 			"Interrupt Cause Tx Queue Min Thresh Count");
4246 
4247 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_desc_min_thresh",
4248 			CTLFLAG_RD, &adapter->stats.icrxdmtc,
4249 			"Interrupt Cause Rx Desc Min Thresh Count");
4250 
4251 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_overrun",
4252 			CTLFLAG_RD, &adapter->stats.icrxoc,
4253 			"Interrupt Cause Receiver Overrun Count");
4254 }
4255 
4256 /**********************************************************************
4257  *
4258  *  This routine provides a way to dump out the adapter eeprom,
4259  *  often a useful debug/service tool. This only dumps the first
4260  *  32 words, stuff that matters is in that extent.
4261  *
4262  **********************************************************************/
4263 static int
4264 em_sysctl_nvm_info(SYSCTL_HANDLER_ARGS)
4265 {
4266 	struct adapter *adapter = (struct adapter *)arg1;
4267 	int error;
4268 	int result;
4269 
4270 	result = -1;
4271 	error = sysctl_handle_int(oidp, &result, 0, req);
4272 
4273 	if (error || !req->newptr)
4274 		return (error);
4275 
4276 	/*
4277 	 * This value will cause a hex dump of the
4278 	 * first 32 16-bit words of the EEPROM to
4279 	 * the screen.
4280 	 */
4281 	if (result == 1)
4282 		em_print_nvm_info(adapter);
4283 
4284 	return (error);
4285 }
4286 
4287 static void
4288 em_print_nvm_info(struct adapter *adapter)
4289 {
4290 	u16 eeprom_data;
4291 	int i, j, row = 0;
4292 
4293 	/* Its a bit crude, but it gets the job done */
4294 	printf("\nInterface EEPROM Dump:\n");
4295 	printf("Offset\n0x0000  ");
4296 	for (i = 0, j = 0; i < 32; i++, j++) {
4297 		if (j == 8) { /* Make the offset block */
4298 			j = 0; ++row;
4299 			printf("\n0x00%x0  ",row);
4300 		}
4301 		e1000_read_nvm(&adapter->hw, i, 1, &eeprom_data);
4302 		printf("%04x ", eeprom_data);
4303 	}
4304 	printf("\n");
4305 }
4306 
4307 static int
4308 em_sysctl_int_delay(SYSCTL_HANDLER_ARGS)
4309 {
4310 	struct em_int_delay_info *info;
4311 	struct adapter *adapter;
4312 	u32 regval;
4313 	int error, usecs, ticks;
4314 
4315 	info = (struct em_int_delay_info *) arg1;
4316 	usecs = info->value;
4317 	error = sysctl_handle_int(oidp, &usecs, 0, req);
4318 	if (error != 0 || req->newptr == NULL)
4319 		return (error);
4320 	if (usecs < 0 || usecs > EM_TICKS_TO_USECS(65535))
4321 		return (EINVAL);
4322 	info->value = usecs;
4323 	ticks = EM_USECS_TO_TICKS(usecs);
4324 	if (info->offset == E1000_ITR)	/* units are 256ns here */
4325 		ticks *= 4;
4326 
4327 	adapter = info->adapter;
4328 
4329 	regval = E1000_READ_OFFSET(&adapter->hw, info->offset);
4330 	regval = (regval & ~0xffff) | (ticks & 0xffff);
4331 	/* Handle a few special cases. */
4332 	switch (info->offset) {
4333 	case E1000_RDTR:
4334 		break;
4335 	case E1000_TIDV:
4336 		if (ticks == 0) {
4337 			adapter->txd_cmd &= ~E1000_TXD_CMD_IDE;
4338 			/* Don't write 0 into the TIDV register. */
4339 			regval++;
4340 		} else
4341 			adapter->txd_cmd |= E1000_TXD_CMD_IDE;
4342 		break;
4343 	}
4344 	E1000_WRITE_OFFSET(&adapter->hw, info->offset, regval);
4345 	return (0);
4346 }
4347 
4348 static void
4349 em_add_int_delay_sysctl(struct adapter *adapter, const char *name,
4350 	const char *description, struct em_int_delay_info *info,
4351 	int offset, int value)
4352 {
4353 	info->adapter = adapter;
4354 	info->offset = offset;
4355 	info->value = value;
4356 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(adapter->dev),
4357 	    SYSCTL_CHILDREN(device_get_sysctl_tree(adapter->dev)),
4358 	    OID_AUTO, name, CTLTYPE_INT|CTLFLAG_RW,
4359 	    info, 0, em_sysctl_int_delay, "I", description);
4360 }
4361 
4362 /*
4363  * Set flow control using sysctl:
4364  * Flow control values:
4365  *      0 - off
4366  *      1 - rx pause
4367  *      2 - tx pause
4368  *      3 - full
4369  */
4370 static int
4371 em_set_flowcntl(SYSCTL_HANDLER_ARGS)
4372 {
4373 	int error;
4374 	static int input = 3; /* default is full */
4375 	struct adapter	*adapter = (struct adapter *) arg1;
4376 
4377 	error = sysctl_handle_int(oidp, &input, 0, req);
4378 
4379 	if ((error) || (req->newptr == NULL))
4380 		return (error);
4381 
4382 	if (input == adapter->fc) /* no change? */
4383 		return (error);
4384 
4385 	switch (input) {
4386 	case e1000_fc_rx_pause:
4387 	case e1000_fc_tx_pause:
4388 	case e1000_fc_full:
4389 	case e1000_fc_none:
4390 		adapter->hw.fc.requested_mode = input;
4391 		adapter->fc = input;
4392 		break;
4393 	default:
4394 		/* Do nothing */
4395 		return (error);
4396 	}
4397 
4398 	adapter->hw.fc.current_mode = adapter->hw.fc.requested_mode;
4399 	e1000_force_mac_fc(&adapter->hw);
4400 	return (error);
4401 }
4402 
4403 /*
4404  * Manage Energy Efficient Ethernet:
4405  * Control values:
4406  *     0/1 - enabled/disabled
4407  */
4408 static int
4409 em_sysctl_eee(SYSCTL_HANDLER_ARGS)
4410 {
4411 	struct adapter *adapter = (struct adapter *) arg1;
4412 	int error, value;
4413 
4414 	value = adapter->hw.dev_spec.ich8lan.eee_disable;
4415 	error = sysctl_handle_int(oidp, &value, 0, req);
4416 	if (error || req->newptr == NULL)
4417 		return (error);
4418 	adapter->hw.dev_spec.ich8lan.eee_disable = (value != 0);
4419 	em_if_init(adapter->ctx);
4420 
4421 	return (0);
4422 }
4423 
4424 static int
4425 em_sysctl_debug_info(SYSCTL_HANDLER_ARGS)
4426 {
4427 	struct adapter *adapter;
4428 	int error;
4429 	int result;
4430 
4431 	result = -1;
4432 	error = sysctl_handle_int(oidp, &result, 0, req);
4433 
4434 	if (error || !req->newptr)
4435 		return (error);
4436 
4437 	if (result == 1) {
4438 		adapter = (struct adapter *) arg1;
4439 		em_print_debug_info(adapter);
4440 	}
4441 
4442 	return (error);
4443 }
4444 
4445 static int
4446 em_get_rs(SYSCTL_HANDLER_ARGS)
4447 {
4448 	struct adapter *adapter = (struct adapter *) arg1;
4449 	int error;
4450 	int result;
4451 
4452 	result = 0;
4453 	error = sysctl_handle_int(oidp, &result, 0, req);
4454 
4455 	if (error || !req->newptr || result != 1)
4456 		return (error);
4457 	em_dump_rs(adapter);
4458 
4459 	return (error);
4460 }
4461 
4462 static void
4463 em_if_debug(if_ctx_t ctx)
4464 {
4465 	em_dump_rs(iflib_get_softc(ctx));
4466 }
4467 
4468 /*
4469  * This routine is meant to be fluid, add whatever is
4470  * needed for debugging a problem.  -jfv
4471  */
4472 static void
4473 em_print_debug_info(struct adapter *adapter)
4474 {
4475 	device_t dev = iflib_get_dev(adapter->ctx);
4476 	struct ifnet *ifp = iflib_get_ifp(adapter->ctx);
4477 	struct tx_ring *txr = &adapter->tx_queues->txr;
4478 	struct rx_ring *rxr = &adapter->rx_queues->rxr;
4479 
4480 	if (if_getdrvflags(ifp) & IFF_DRV_RUNNING)
4481 		printf("Interface is RUNNING ");
4482 	else
4483 		printf("Interface is NOT RUNNING\n");
4484 
4485 	if (if_getdrvflags(ifp) & IFF_DRV_OACTIVE)
4486 		printf("and INACTIVE\n");
4487 	else
4488 		printf("and ACTIVE\n");
4489 
4490 	for (int i = 0; i < adapter->tx_num_queues; i++, txr++) {
4491 		device_printf(dev, "TX Queue %d ------\n", i);
4492 		device_printf(dev, "hw tdh = %d, hw tdt = %d\n",
4493 			E1000_READ_REG(&adapter->hw, E1000_TDH(i)),
4494 			E1000_READ_REG(&adapter->hw, E1000_TDT(i)));
4495 
4496 	}
4497 	for (int j=0; j < adapter->rx_num_queues; j++, rxr++) {
4498 		device_printf(dev, "RX Queue %d ------\n", j);
4499 		device_printf(dev, "hw rdh = %d, hw rdt = %d\n",
4500 			E1000_READ_REG(&adapter->hw, E1000_RDH(j)),
4501 			E1000_READ_REG(&adapter->hw, E1000_RDT(j)));
4502 	}
4503 }
4504 
4505 /*
4506  * 82574 only:
4507  * Write a new value to the EEPROM increasing the number of MSIX
4508  * vectors from 3 to 5, for proper multiqueue support.
4509  */
4510 static void
4511 em_enable_vectors_82574(if_ctx_t ctx)
4512 {
4513 	struct adapter *adapter = iflib_get_softc(ctx);
4514 	struct e1000_hw *hw = &adapter->hw;
4515 	device_t dev = iflib_get_dev(ctx);
4516 	u16 edata;
4517 
4518 	e1000_read_nvm(hw, EM_NVM_PCIE_CTRL, 1, &edata);
4519 	printf("Current cap: %#06x\n", edata);
4520 	if (((edata & EM_NVM_MSIX_N_MASK) >> EM_NVM_MSIX_N_SHIFT) != 4) {
4521 		device_printf(dev, "Writing to eeprom: increasing "
4522 		    "reported MSIX vectors from 3 to 5...\n");
4523 		edata &= ~(EM_NVM_MSIX_N_MASK);
4524 		edata |= 4 << EM_NVM_MSIX_N_SHIFT;
4525 		e1000_write_nvm(hw, EM_NVM_PCIE_CTRL, 1, &edata);
4526 		e1000_update_nvm_checksum(hw);
4527 		device_printf(dev, "Writing to eeprom: done\n");
4528 	}
4529 }
4530