xref: /freebsd/sys/dev/e1000/if_em.c (revision ddba0402f443ad5f70ea315ede04aa7d6fa53691)
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_COPPER,	"Intel(R) PRO/1000 Network Connection"),
108 	PVID(0x8086, E1000_DEV_ID_82572EI_FIBER,	"Intel(R) PRO/1000 Network Connection"),
109 	PVID(0x8086, E1000_DEV_ID_82572EI_SERDES,	"Intel(R) PRO/1000 Network Connection"),
110 	PVID(0x8086, E1000_DEV_ID_82573E,		"Intel(R) PRO/1000 Network Connection"),
111 	PVID(0x8086, E1000_DEV_ID_82573E_IAMT,	"Intel(R) PRO/1000 Network Connection"),
112 	PVID(0x8086, E1000_DEV_ID_82573L,		"Intel(R) PRO/1000 Network Connection"),
113 	PVID(0x8086, E1000_DEV_ID_82583V,		"Intel(R) PRO/1000 Network Connection"),
114 	PVID(0x8086, E1000_DEV_ID_80003ES2LAN_COPPER_SPT, "Intel(R) PRO/1000 Network Connection"),
115 	PVID(0x8086, E1000_DEV_ID_80003ES2LAN_SERDES_SPT, "Intel(R) PRO/1000 Network Connection"),
116 	PVID(0x8086, E1000_DEV_ID_80003ES2LAN_COPPER_DPT, "Intel(R) PRO/1000 Network Connection"),
117 	PVID(0x8086, E1000_DEV_ID_80003ES2LAN_SERDES_DPT, "Intel(R) PRO/1000 Network Connection"),
118 	PVID(0x8086, E1000_DEV_ID_ICH8_IGP_M_AMT,	"Intel(R) PRO/1000 Network Connection"),
119 	PVID(0x8086, E1000_DEV_ID_ICH8_IGP_AMT,	"Intel(R) PRO/1000 Network Connection"),
120 	PVID(0x8086, E1000_DEV_ID_ICH8_IGP_C,	"Intel(R) PRO/1000 Network Connection"),
121 	PVID(0x8086, E1000_DEV_ID_ICH8_IFE,	"Intel(R) PRO/1000 Network Connection"),
122 	PVID(0x8086, E1000_DEV_ID_ICH8_IFE_GT,	"Intel(R) PRO/1000 Network Connection"),
123 	PVID(0x8086, E1000_DEV_ID_ICH8_IFE_G,	"Intel(R) PRO/1000 Network Connection"),
124 	PVID(0x8086, E1000_DEV_ID_ICH8_IGP_M,	"Intel(R) PRO/1000 Network Connection"),
125 	PVID(0x8086, E1000_DEV_ID_ICH8_82567V_3,	"Intel(R) PRO/1000 Network Connection"),
126 	PVID(0x8086, E1000_DEV_ID_ICH9_IGP_M_AMT,	"Intel(R) PRO/1000 Network Connection"),
127 	PVID(0x8086, E1000_DEV_ID_ICH9_IGP_AMT,	"Intel(R) PRO/1000 Network Connection"),
128 	PVID(0x8086, E1000_DEV_ID_ICH9_IGP_C,	"Intel(R) PRO/1000 Network Connection"),
129 	PVID(0x8086, E1000_DEV_ID_ICH9_IGP_M,	"Intel(R) PRO/1000 Network Connection"),
130 	PVID(0x8086, E1000_DEV_ID_ICH9_IGP_M_V,	"Intel(R) PRO/1000 Network Connection"),
131 	PVID(0x8086, E1000_DEV_ID_ICH9_IFE,	"Intel(R) PRO/1000 Network Connection"),
132 	PVID(0x8086, E1000_DEV_ID_ICH9_IFE_GT,	"Intel(R) PRO/1000 Network Connection"),
133 	PVID(0x8086, E1000_DEV_ID_ICH9_IFE_G,	"Intel(R) PRO/1000 Network Connection"),
134 	PVID(0x8086, E1000_DEV_ID_ICH9_BM,		"Intel(R) PRO/1000 Network Connection"),
135 	PVID(0x8086, E1000_DEV_ID_82574L,		"Intel(R) PRO/1000 Network Connection"),
136 	PVID(0x8086, E1000_DEV_ID_82574LA,		"Intel(R) PRO/1000 Network Connection"),
137 	PVID(0x8086, E1000_DEV_ID_ICH10_R_BM_LM,	"Intel(R) PRO/1000 Network Connection"),
138 	PVID(0x8086, E1000_DEV_ID_ICH10_R_BM_LF,	"Intel(R) PRO/1000 Network Connection"),
139 	PVID(0x8086, E1000_DEV_ID_ICH10_R_BM_V,	"Intel(R) PRO/1000 Network Connection"),
140 	PVID(0x8086, E1000_DEV_ID_ICH10_D_BM_LM,	"Intel(R) PRO/1000 Network Connection"),
141 	PVID(0x8086, E1000_DEV_ID_ICH10_D_BM_LF,	"Intel(R) PRO/1000 Network Connection"),
142 	PVID(0x8086, E1000_DEV_ID_ICH10_D_BM_V,	"Intel(R) PRO/1000 Network Connection"),
143 	PVID(0x8086, E1000_DEV_ID_PCH_M_HV_LM,	"Intel(R) PRO/1000 Network Connection"),
144 	PVID(0x8086, E1000_DEV_ID_PCH_M_HV_LC,	"Intel(R) PRO/1000 Network Connection"),
145 	PVID(0x8086, E1000_DEV_ID_PCH_D_HV_DM,	"Intel(R) PRO/1000 Network Connection"),
146 	PVID(0x8086, E1000_DEV_ID_PCH_D_HV_DC,	"Intel(R) PRO/1000 Network Connection"),
147 	PVID(0x8086, E1000_DEV_ID_PCH2_LV_LM,	"Intel(R) PRO/1000 Network Connection"),
148 	PVID(0x8086, E1000_DEV_ID_PCH2_LV_V,	"Intel(R) PRO/1000 Network Connection"),
149 	PVID(0x8086, E1000_DEV_ID_PCH_LPT_I217_LM,	"Intel(R) PRO/1000 Network Connection"),
150 	PVID(0x8086, E1000_DEV_ID_PCH_LPT_I217_V,	"Intel(R) PRO/1000 Network Connection"),
151 	PVID(0x8086, E1000_DEV_ID_PCH_LPTLP_I218_LM, "Intel(R) PRO/1000 Network Connection"),
152 	PVID(0x8086, E1000_DEV_ID_PCH_LPTLP_I218_V, "Intel(R) PRO/1000 Network Connection"),
153 	PVID(0x8086, E1000_DEV_ID_PCH_I218_LM2,	"Intel(R) PRO/1000 Network Connection"),
154 	PVID(0x8086, E1000_DEV_ID_PCH_I218_V2,	"Intel(R) PRO/1000 Network Connection"),
155 	PVID(0x8086, E1000_DEV_ID_PCH_I218_LM3,	"Intel(R) PRO/1000 Network Connection"),
156 	PVID(0x8086, E1000_DEV_ID_PCH_I218_V3,	"Intel(R) PRO/1000 Network Connection"),
157 	PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_LM, "Intel(R) PRO/1000 Network Connection"),
158 	PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_V,  "Intel(R) PRO/1000 Network Connection"),
159 	PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_LM2, "Intel(R) PRO/1000 Network Connection"),
160 	PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_V2, "Intel(R) PRO/1000 Network Connection"),
161 	PVID(0x8086, E1000_DEV_ID_PCH_LBG_I219_LM3, "Intel(R) PRO/1000 Network Connection"),
162 	/* required last entry */
163 	PVID_END
164 };
165 
166 static pci_vendor_info_t igb_vendor_info_array[] =
167 {
168 	/* Intel(R) PRO/1000 Network Connection - em */
169 	PVID(0x8086, E1000_DEV_ID_82575EB_COPPER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
170 	PVID(0x8086, E1000_DEV_ID_82575EB_FIBER_SERDES, "Intel(R) PRO/1000 PCI-Express Network Driver"),
171 	PVID(0x8086, E1000_DEV_ID_82575GB_QUAD_COPPER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
172 	PVID(0x8086, E1000_DEV_ID_82576, "Intel(R) PRO/1000 PCI-Express Network Driver"),
173 	PVID(0x8086, E1000_DEV_ID_82576_NS, "Intel(R) PRO/1000 PCI-Express Network Driver"),
174 	PVID(0x8086, E1000_DEV_ID_82576_NS_SERDES, "Intel(R) PRO/1000 PCI-Express Network Driver"),
175 	PVID(0x8086, E1000_DEV_ID_82576_FIBER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
176 	PVID(0x8086, E1000_DEV_ID_82576_SERDES, "Intel(R) PRO/1000 PCI-Express Network Driver"),
177 	PVID(0x8086, E1000_DEV_ID_82576_SERDES_QUAD, "Intel(R) PRO/1000 PCI-Express Network Driver"),
178 	PVID(0x8086, E1000_DEV_ID_82576_QUAD_COPPER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
179 	PVID(0x8086, E1000_DEV_ID_82576_QUAD_COPPER_ET2, "Intel(R) PRO/1000 PCI-Express Network Driver"),
180 	PVID(0x8086, E1000_DEV_ID_82576_VF, "Intel(R) PRO/1000 PCI-Express Network Driver"),
181 	PVID(0x8086, E1000_DEV_ID_82580_COPPER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
182 	PVID(0x8086, E1000_DEV_ID_82580_FIBER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
183 	PVID(0x8086, E1000_DEV_ID_82580_SERDES, "Intel(R) PRO/1000 PCI-Express Network Driver"),
184 	PVID(0x8086, E1000_DEV_ID_82580_SGMII, "Intel(R) PRO/1000 PCI-Express Network Driver"),
185 	PVID(0x8086, E1000_DEV_ID_82580_COPPER_DUAL, "Intel(R) PRO/1000 PCI-Express Network Driver"),
186 	PVID(0x8086, E1000_DEV_ID_82580_QUAD_FIBER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
187 	PVID(0x8086, E1000_DEV_ID_DH89XXCC_SERDES, "Intel(R) PRO/1000 PCI-Express Network Driver"),
188 	PVID(0x8086, E1000_DEV_ID_DH89XXCC_SGMII, "Intel(R) PRO/1000 PCI-Express Network Driver"),
189 	PVID(0x8086, E1000_DEV_ID_DH89XXCC_SFP, "Intel(R) PRO/1000 PCI-Express Network Driver"),
190 	PVID(0x8086, E1000_DEV_ID_DH89XXCC_BACKPLANE, "Intel(R) PRO/1000 PCI-Express Network Driver"),
191 	PVID(0x8086, E1000_DEV_ID_I350_COPPER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
192 	PVID(0x8086, E1000_DEV_ID_I350_FIBER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
193 	PVID(0x8086, E1000_DEV_ID_I350_SERDES, "Intel(R) PRO/1000 PCI-Express Network Driver"),
194 	PVID(0x8086, E1000_DEV_ID_I350_SGMII, "Intel(R) PRO/1000 PCI-Express Network Driver"),
195 	PVID(0x8086, E1000_DEV_ID_I350_VF, "Intel(R) PRO/1000 PCI-Express Network Driver"),
196 	PVID(0x8086, E1000_DEV_ID_I210_COPPER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
197 	PVID(0x8086, E1000_DEV_ID_I210_COPPER_IT, "Intel(R) PRO/1000 PCI-Express Network Driver"),
198 	PVID(0x8086, E1000_DEV_ID_I210_COPPER_OEM1, "Intel(R) PRO/1000 PCI-Express Network Driver"),
199 	PVID(0x8086, E1000_DEV_ID_I210_COPPER_FLASHLESS, "Intel(R) PRO/1000 PCI-Express Network Driver"),
200 	PVID(0x8086, E1000_DEV_ID_I210_SERDES_FLASHLESS, "Intel(R) PRO/1000 PCI-Express Network Driver"),
201 	PVID(0x8086, E1000_DEV_ID_I210_FIBER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
202 	PVID(0x8086, E1000_DEV_ID_I210_SERDES, "Intel(R) PRO/1000 PCI-Express Network Driver"),
203 	PVID(0x8086, E1000_DEV_ID_I210_SGMII, "Intel(R) PRO/1000 PCI-Express Network Driver"),
204 	PVID(0x8086, E1000_DEV_ID_I211_COPPER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
205 	PVID(0x8086, E1000_DEV_ID_I354_BACKPLANE_1GBPS, "Intel(R) PRO/1000 PCI-Express Network Driver"),
206 	PVID(0x8086, E1000_DEV_ID_I354_BACKPLANE_2_5GBPS, "Intel(R) PRO/1000 PCI-Express Network Driver"),
207 	PVID(0x8086, E1000_DEV_ID_I354_SGMII, "Intel(R) PRO/1000 PCI-Express Network Driver"),
208 	/* required last entry */
209 	PVID_END
210 };
211 
212 /*********************************************************************
213  *  Function prototypes
214  *********************************************************************/
215 static void     *em_register(device_t dev);
216 static void     *igb_register(device_t dev);
217 static int	em_if_attach_pre(if_ctx_t ctx);
218 static int	em_if_attach_post(if_ctx_t ctx);
219 static int	em_if_detach(if_ctx_t ctx);
220 static int	em_if_shutdown(if_ctx_t ctx);
221 static int	em_if_suspend(if_ctx_t ctx);
222 static int	em_if_resume(if_ctx_t ctx);
223 
224 static int      em_if_tx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int ntxqs, int ntxqsets);
225 static int      em_if_rx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int nrxqs, int nrxqsets);
226 static void     em_if_queues_free(if_ctx_t ctx);
227 
228 static uint64_t	em_if_get_counter(if_ctx_t, ift_counter);
229 static void	em_if_init(if_ctx_t ctx);
230 static void	em_if_stop(if_ctx_t ctx);
231 static void	em_if_media_status(if_ctx_t, struct ifmediareq *);
232 static int	em_if_media_change(if_ctx_t ctx);
233 static int      em_if_mtu_set(if_ctx_t ctx, uint32_t mtu);
234 static void     em_if_timer(if_ctx_t ctx, uint16_t qid);
235 static void     em_if_vlan_register(if_ctx_t ctx, u16 vtag);
236 static void     em_if_vlan_unregister(if_ctx_t ctx, u16 vtag);
237 
238 static void	em_identify_hardware(if_ctx_t ctx);
239 static int	em_allocate_pci_resources(if_ctx_t ctx);
240 static void	em_free_pci_resources(if_ctx_t ctx);
241 static void	em_reset(if_ctx_t ctx);
242 static int	em_setup_interface(if_ctx_t ctx);
243 static int      em_setup_msix(if_ctx_t ctx);
244 
245 static void	em_initialize_transmit_unit(if_ctx_t ctx);
246 static void	em_initialize_receive_unit(if_ctx_t ctx);
247 
248 static void	em_if_enable_intr(if_ctx_t ctx);
249 static void	em_if_disable_intr(if_ctx_t ctx);
250 static int      em_if_queue_intr_enable(if_ctx_t ctx, uint16_t rxqid);
251 static void     em_if_multi_set(if_ctx_t ctx);
252 static void     em_if_update_admin_status(if_ctx_t ctx);
253 static void	em_update_stats_counters(struct adapter *);
254 static void	em_add_hw_stats(struct adapter *adapter);
255 static int	em_if_set_promisc(if_ctx_t ctx, int flags);
256 static void	em_setup_vlan_hw_support(struct adapter *);
257 static int	em_sysctl_nvm_info(SYSCTL_HANDLER_ARGS);
258 static void	em_print_nvm_info(struct adapter *);
259 static int	em_sysctl_debug_info(SYSCTL_HANDLER_ARGS);
260 static void	em_print_debug_info(struct adapter *);
261 static int 	em_is_valid_ether_addr(u8 *);
262 static int	em_sysctl_int_delay(SYSCTL_HANDLER_ARGS);
263 static void	em_add_int_delay_sysctl(struct adapter *, const char *,
264 		    const char *, struct em_int_delay_info *, int, int);
265 /* Management and WOL Support */
266 static void	em_init_manageability(struct adapter *);
267 static void	em_release_manageability(struct adapter *);
268 static void     em_get_hw_control(struct adapter *);
269 static void     em_release_hw_control(struct adapter *);
270 static void	em_get_wakeup(if_ctx_t ctx);
271 static void     em_enable_wakeup(if_ctx_t ctx);
272 static int	em_enable_phy_wakeup(struct adapter *);
273 static void	em_disable_aspm(struct adapter *);
274 
275 int             em_intr(void *arg);
276 static void     em_disable_promisc(if_ctx_t ctx);
277 
278 /* MSIX handlers */
279 static int      em_if_msix_intr_assign(if_ctx_t, int);
280 static int	em_msix_link(void *);
281 static void	em_handle_link(void *context);
282 
283 static void	em_enable_vectors_82574(if_ctx_t);
284 
285 static void	em_set_sysctl_value(struct adapter *, const char *,
286 		    const char *, int *, int);
287 static int	em_set_flowcntl(SYSCTL_HANDLER_ARGS);
288 static int	em_sysctl_eee(SYSCTL_HANDLER_ARGS);
289 static void     em_if_led_func(if_ctx_t ctx, int onoff);
290 
291 static void	em_init_tx_ring(struct em_tx_queue *que);
292 static int	em_get_regs(SYSCTL_HANDLER_ARGS);
293 
294 static void	lem_smartspeed(struct adapter *adapter);
295 static void	igb_configure_queues(struct adapter *adapter);
296 
297 
298 /*********************************************************************
299  *  FreeBSD Device Interface Entry Points
300  *********************************************************************/
301 static device_method_t em_methods[] = {
302 	/* Device interface */
303   DEVMETHOD(device_register, em_register),
304   DEVMETHOD(device_probe, iflib_device_probe),
305   DEVMETHOD(device_attach, iflib_device_attach),
306   DEVMETHOD(device_detach, iflib_device_detach),
307   DEVMETHOD(device_shutdown, iflib_device_shutdown),
308   DEVMETHOD(device_suspend, iflib_device_suspend),
309   DEVMETHOD(device_resume, iflib_device_resume),
310   DEVMETHOD_END
311 };
312 
313 static device_method_t igb_methods[] = {
314 	/* Device interface */
315   DEVMETHOD(device_register, igb_register),
316   DEVMETHOD(device_probe, iflib_device_probe),
317   DEVMETHOD(device_attach, iflib_device_attach),
318   DEVMETHOD(device_detach, iflib_device_detach),
319   DEVMETHOD(device_shutdown, iflib_device_shutdown),
320   DEVMETHOD(device_suspend, iflib_device_suspend),
321   DEVMETHOD(device_resume, iflib_device_resume),
322   DEVMETHOD_END
323 };
324 
325 
326 static driver_t em_driver = {
327 	"em", em_methods, sizeof(struct adapter),
328 };
329 
330 static devclass_t em_devclass;
331 DRIVER_MODULE(em, pci, em_driver, em_devclass, 0, 0);
332 
333 MODULE_DEPEND(em, pci, 1, 1, 1);
334 MODULE_DEPEND(em, ether, 1, 1, 1);
335 MODULE_DEPEND(em, iflib, 1, 1, 1);
336 
337 static driver_t igb_driver = {
338 	"igb", igb_methods, sizeof(struct adapter),
339 };
340 
341 static devclass_t igb_devclass;
342 DRIVER_MODULE(igb, pci, igb_driver, igb_devclass, 0, 0);
343 
344 MODULE_DEPEND(igb, pci, 1, 1, 1);
345 MODULE_DEPEND(igb, ether, 1, 1, 1);
346 MODULE_DEPEND(igb, iflib, 1, 1, 1);
347 
348 
349 static device_method_t em_if_methods[] = {
350         DEVMETHOD(ifdi_attach_pre, em_if_attach_pre),
351   	DEVMETHOD(ifdi_attach_post, em_if_attach_post),
352 	DEVMETHOD(ifdi_detach, em_if_detach),
353 	DEVMETHOD(ifdi_shutdown, em_if_shutdown),
354 	DEVMETHOD(ifdi_suspend, em_if_suspend),
355 	DEVMETHOD(ifdi_resume, em_if_resume),
356 	DEVMETHOD(ifdi_init, em_if_init),
357 	DEVMETHOD(ifdi_stop, em_if_stop),
358 	DEVMETHOD(ifdi_msix_intr_assign, em_if_msix_intr_assign),
359 	DEVMETHOD(ifdi_intr_enable, em_if_enable_intr),
360 	DEVMETHOD(ifdi_intr_disable, em_if_disable_intr),
361 	DEVMETHOD(ifdi_tx_queues_alloc, em_if_tx_queues_alloc),
362 	DEVMETHOD(ifdi_rx_queues_alloc, em_if_rx_queues_alloc),
363 	DEVMETHOD(ifdi_queues_free, em_if_queues_free),
364 	DEVMETHOD(ifdi_update_admin_status, em_if_update_admin_status),
365 	DEVMETHOD(ifdi_multi_set, em_if_multi_set),
366 	DEVMETHOD(ifdi_media_status, em_if_media_status),
367 	DEVMETHOD(ifdi_media_change, em_if_media_change),
368 	DEVMETHOD(ifdi_mtu_set, em_if_mtu_set),
369 	DEVMETHOD(ifdi_promisc_set, em_if_set_promisc),
370 	DEVMETHOD(ifdi_timer, em_if_timer),
371 	DEVMETHOD(ifdi_vlan_register, em_if_vlan_register),
372 	DEVMETHOD(ifdi_vlan_unregister, em_if_vlan_unregister),
373 	DEVMETHOD(ifdi_get_counter, em_if_get_counter),
374 	DEVMETHOD(ifdi_led_func, em_if_led_func),
375 	DEVMETHOD(ifdi_queue_intr_enable, em_if_queue_intr_enable),
376 	DEVMETHOD_END
377 };
378 
379   /*
380  * note that if (adapter->msix_mem) is replaced by:
381  * if (adapter->intr_type == IFLIB_INTR_MSIX)
382  */
383 static driver_t em_if_driver = {
384   "em_if", em_if_methods, sizeof(struct adapter)
385 };
386 
387 /*********************************************************************
388  *  Tunable default values.
389  *********************************************************************/
390 
391 #define EM_TICKS_TO_USECS(ticks)	((1024 * (ticks) + 500) / 1000)
392 #define EM_USECS_TO_TICKS(usecs)	((1000 * (usecs) + 512) / 1024)
393 #define M_TSO_LEN			66
394 
395 #define MAX_INTS_PER_SEC	8000
396 #define DEFAULT_ITR		(1000000000/(MAX_INTS_PER_SEC * 256))
397 
398 /* Allow common code without TSO */
399 #ifndef CSUM_TSO
400 #define CSUM_TSO	0
401 #endif
402 
403 #define TSO_WORKAROUND	4
404 
405 static SYSCTL_NODE(_hw, OID_AUTO, em, CTLFLAG_RD, 0, "EM driver parameters");
406 
407 static int em_disable_crc_stripping = 0;
408 SYSCTL_INT(_hw_em, OID_AUTO, disable_crc_stripping, CTLFLAG_RDTUN,
409     &em_disable_crc_stripping, 0, "Disable CRC Stripping");
410 
411 static int em_tx_int_delay_dflt = EM_TICKS_TO_USECS(EM_TIDV);
412 static int em_rx_int_delay_dflt = EM_TICKS_TO_USECS(EM_RDTR);
413 SYSCTL_INT(_hw_em, OID_AUTO, tx_int_delay, CTLFLAG_RDTUN, &em_tx_int_delay_dflt,
414     0, "Default transmit interrupt delay in usecs");
415 SYSCTL_INT(_hw_em, OID_AUTO, rx_int_delay, CTLFLAG_RDTUN, &em_rx_int_delay_dflt,
416     0, "Default receive interrupt delay in usecs");
417 
418 static int em_tx_abs_int_delay_dflt = EM_TICKS_TO_USECS(EM_TADV);
419 static int em_rx_abs_int_delay_dflt = EM_TICKS_TO_USECS(EM_RADV);
420 SYSCTL_INT(_hw_em, OID_AUTO, tx_abs_int_delay, CTLFLAG_RDTUN,
421     &em_tx_abs_int_delay_dflt, 0,
422     "Default transmit interrupt delay limit in usecs");
423 SYSCTL_INT(_hw_em, OID_AUTO, rx_abs_int_delay, CTLFLAG_RDTUN,
424     &em_rx_abs_int_delay_dflt, 0,
425     "Default receive interrupt delay limit in usecs");
426 
427 static int em_smart_pwr_down = FALSE;
428 SYSCTL_INT(_hw_em, OID_AUTO, smart_pwr_down, CTLFLAG_RDTUN, &em_smart_pwr_down,
429     0, "Set to true to leave smart power down enabled on newer adapters");
430 
431 /* Controls whether promiscuous also shows bad packets */
432 static int em_debug_sbp = TRUE;
433 SYSCTL_INT(_hw_em, OID_AUTO, sbp, CTLFLAG_RDTUN, &em_debug_sbp, 0,
434     "Show bad packets in promiscuous mode");
435 
436 /* How many packets rxeof tries to clean at a time */
437 static int em_rx_process_limit = 100;
438 SYSCTL_INT(_hw_em, OID_AUTO, rx_process_limit, CTLFLAG_RDTUN,
439     &em_rx_process_limit, 0,
440     "Maximum number of received packets to process "
441     "at a time, -1 means unlimited");
442 
443 /* Energy efficient ethernet - default to OFF */
444 static int eee_setting = 1;
445 SYSCTL_INT(_hw_em, OID_AUTO, eee_setting, CTLFLAG_RDTUN, &eee_setting, 0,
446     "Enable Energy Efficient Ethernet");
447 
448 /*
449 ** Tuneable Interrupt rate
450 */
451 static int em_max_interrupt_rate = 8000;
452 SYSCTL_INT(_hw_em, OID_AUTO, max_interrupt_rate, CTLFLAG_RDTUN,
453     &em_max_interrupt_rate, 0, "Maximum interrupts per second");
454 
455 
456 
457 /* Global used in WOL setup with multiport cards */
458 static int global_quad_port_a = 0;
459 
460 extern struct if_txrx igb_txrx;
461 extern struct if_txrx em_txrx;
462 extern struct if_txrx lem_txrx;
463 
464 static struct if_shared_ctx em_sctx_init = {
465     	.isc_magic = IFLIB_MAGIC,
466 	.isc_q_align = PAGE_SIZE,
467 	.isc_tx_maxsize = EM_TSO_SIZE,
468 	.isc_tx_maxsegsize = PAGE_SIZE,
469 	.isc_rx_maxsize = MJUM9BYTES,
470 	.isc_rx_nsegments = 1,
471 	.isc_rx_maxsegsize = MJUM9BYTES,
472 	.isc_nfl = 1,
473 	.isc_nrxqs = 1,
474 	.isc_ntxqs = 1,
475 	.isc_admin_intrcnt = 1,
476 	.isc_vendor_info = em_vendor_info_array,
477 	.isc_driver_version = em_driver_version,
478 	.isc_driver = &em_if_driver,
479 	.isc_flags = IFLIB_NEED_SCRATCH | IFLIB_TSO_INIT_IP,
480 
481 	.isc_nrxd_min = {EM_MIN_RXD},
482 	.isc_ntxd_min = {EM_MIN_TXD},
483 	.isc_nrxd_max = {EM_MAX_RXD},
484 	.isc_ntxd_max = {EM_MAX_TXD},
485 	.isc_nrxd_default = {EM_DEFAULT_RXD},
486 	.isc_ntxd_default = {EM_DEFAULT_TXD},
487 };
488 
489 if_shared_ctx_t em_sctx = &em_sctx_init;
490 
491 
492 static struct if_shared_ctx igb_sctx_init = {
493     	.isc_magic = IFLIB_MAGIC,
494 	.isc_q_align = PAGE_SIZE,
495 	.isc_tx_maxsize = EM_TSO_SIZE,
496 	.isc_tx_maxsegsize = PAGE_SIZE,
497 	.isc_rx_maxsize = MJUM9BYTES,
498 	.isc_rx_nsegments = 1,
499 	.isc_rx_maxsegsize = MJUM9BYTES,
500 	.isc_nfl = 1,
501 	.isc_nrxqs = 1,
502 	.isc_ntxqs = 1,
503 	.isc_admin_intrcnt = 1,
504 	.isc_vendor_info = igb_vendor_info_array,
505 	.isc_driver_version = em_driver_version,
506 	.isc_driver = &em_if_driver,
507 	.isc_flags = IFLIB_NEED_SCRATCH | IFLIB_TSO_INIT_IP,
508 
509 	.isc_nrxd_min = {EM_MIN_RXD},
510 	.isc_ntxd_min = {EM_MIN_TXD},
511 	.isc_nrxd_max = {EM_MAX_RXD},
512 	.isc_ntxd_max = {EM_MAX_TXD},
513 	.isc_nrxd_default = {EM_DEFAULT_RXD},
514 	.isc_ntxd_default = {EM_DEFAULT_TXD},
515 };
516 
517 if_shared_ctx_t igb_sctx = &igb_sctx_init;
518 
519 /*****************************************************************
520  *
521  * Dump Registers
522  *
523  ****************************************************************/
524 #define IGB_REGS_LEN 739
525 
526 static int em_get_regs(SYSCTL_HANDLER_ARGS)
527 {
528 	struct adapter *adapter = (struct adapter *)arg1;
529 	struct e1000_hw *hw = &adapter->hw;
530 
531 	struct sbuf *sb;
532 	u32 *regs_buff = (u32 *)malloc(sizeof(u32) * IGB_REGS_LEN, M_DEVBUF, M_NOWAIT);
533 	int rc;
534 
535 	memset(regs_buff, 0, IGB_REGS_LEN * sizeof(u32));
536 
537 	rc = sysctl_wire_old_buffer(req, 0);
538 	MPASS(rc == 0);
539 	if (rc != 0)
540 	  return (rc);
541 
542 	sb = sbuf_new_for_sysctl(NULL, NULL, 32*400, req);
543 	MPASS(sb != NULL);
544 	if (sb == NULL)
545 		return (ENOMEM);
546 
547 	/* General Registers */
548 	regs_buff[0] = E1000_READ_REG(hw, E1000_CTRL);
549 	regs_buff[1] = E1000_READ_REG(hw, E1000_STATUS);
550 	regs_buff[2] = E1000_READ_REG(hw, E1000_CTRL_EXT);
551 	regs_buff[3] = E1000_READ_REG(hw, E1000_ICR);
552 	regs_buff[4] = E1000_READ_REG(hw, E1000_RCTL);
553 	regs_buff[5] = E1000_READ_REG(hw, E1000_RDLEN(0));
554 	regs_buff[6] = E1000_READ_REG(hw, E1000_RDH(0));
555 	regs_buff[7] = E1000_READ_REG(hw, E1000_RDT(0));
556 	regs_buff[8] = E1000_READ_REG(hw, E1000_RXDCTL(0));
557 	regs_buff[9] = E1000_READ_REG(hw, E1000_RDBAL(0));
558 	regs_buff[10] = E1000_READ_REG(hw, E1000_RDBAH(0));
559 	regs_buff[11] = E1000_READ_REG(hw, E1000_TCTL);
560 	regs_buff[12] = E1000_READ_REG(hw, E1000_TDBAL(0));
561 	regs_buff[13] = E1000_READ_REG(hw, E1000_TDBAH(0));
562 	regs_buff[14] = E1000_READ_REG(hw, E1000_TDLEN(0));
563 	regs_buff[15] = E1000_READ_REG(hw, E1000_TDH(0));
564 	regs_buff[16] = E1000_READ_REG(hw, E1000_TDT(0));
565 	regs_buff[17] = E1000_READ_REG(hw, E1000_TXDCTL(0));
566 	regs_buff[18] = E1000_READ_REG(hw, E1000_TDFH);
567 	regs_buff[19] = E1000_READ_REG(hw, E1000_TDFT);
568 	regs_buff[20] = E1000_READ_REG(hw, E1000_TDFHS);
569 	regs_buff[21] = E1000_READ_REG(hw, E1000_TDFPC);
570 
571 	sbuf_printf(sb, "General Registers\n");
572 	sbuf_printf(sb, "\tCTRL\t %08x\n", regs_buff[0]);
573 	sbuf_printf(sb, "\tSTATUS\t %08x\n", regs_buff[1]);
574 	sbuf_printf(sb, "\tCTRL_EXIT\t %08x\n\n", regs_buff[2]);
575 
576 	sbuf_printf(sb, "Interrupt Registers\n");
577 	sbuf_printf(sb, "\tICR\t %08x\n\n", regs_buff[3]);
578 
579 	sbuf_printf(sb, "RX Registers\n");
580 	sbuf_printf(sb, "\tRCTL\t %08x\n", regs_buff[4]);
581 	sbuf_printf(sb, "\tRDLEN\t %08x\n", regs_buff[5]);
582 	sbuf_printf(sb, "\tRDH\t %08x\n", regs_buff[6]);
583 	sbuf_printf(sb, "\tRDT\t %08x\n", regs_buff[7]);
584 	sbuf_printf(sb, "\tRXDCTL\t %08x\n", regs_buff[8]);
585 	sbuf_printf(sb, "\tRDBAL\t %08x\n", regs_buff[9]);
586 	sbuf_printf(sb, "\tRDBAH\t %08x\n\n", regs_buff[10]);
587 
588 	sbuf_printf(sb, "TX Registers\n");
589 	sbuf_printf(sb, "\tTCTL\t %08x\n", regs_buff[11]);
590 	sbuf_printf(sb, "\tTDBAL\t %08x\n", regs_buff[12]);
591 	sbuf_printf(sb, "\tTDBAH\t %08x\n", regs_buff[13]);
592 	sbuf_printf(sb, "\tTDLEN\t %08x\n", regs_buff[14]);
593 	sbuf_printf(sb, "\tTDH\t %08x\n", regs_buff[15]);
594 	sbuf_printf(sb, "\tTDT\t %08x\n", regs_buff[16]);
595 	sbuf_printf(sb, "\tTXDCTL\t %08x\n", regs_buff[17]);
596 	sbuf_printf(sb, "\tTDFH\t %08x\n", regs_buff[18]);
597 	sbuf_printf(sb, "\tTDFT\t %08x\n", regs_buff[19]);
598 	sbuf_printf(sb, "\tTDFHS\t %08x\n", regs_buff[20]);
599 	sbuf_printf(sb, "\tTDFPC\t %08x\n\n", regs_buff[21]);
600 
601 #ifdef DUMP_DESCS
602 	{
603 		if_softc_ctx_t scctx = adapter->shared;
604 		struct rx_ring *rxr = &rx_que->rxr;
605 		struct tx_ring *txr = &tx_que->txr;
606 		int ntxd = scctx->isc_ntxd[0];
607 		int nrxd = scctx->isc_nrxd[0];
608 		int j;
609 
610 	for (j = 0; j < nrxd; j++) {
611 		u32 staterr = le32toh(rxr->rx_base[j].wb.upper.status_error);
612 		u32 length =  le32toh(rxr->rx_base[j].wb.upper.length);
613 		sbuf_printf(sb, "\tReceive Descriptor Address %d: %08" PRIx64 "  Error:%d  Length:%d\n", j, rxr->rx_base[j].read.buffer_addr, staterr, length);
614 	}
615 
616 	for (j = 0; j < min(ntxd, 256); j++) {
617 		struct em_txbuffer *buf = &txr->tx_buffers[j];
618 		unsigned int *ptr = (unsigned int *)&txr->tx_base[j];
619 
620 		sbuf_printf(sb, "\tTXD[%03d] [0]: %08x [1]: %08x [2]: %08x [3]: %08x  eop: %d DD=%d\n",
621 			    j, ptr[0], ptr[1], ptr[2], ptr[3], buf->eop,
622 			    buf->eop != -1 ? txr->tx_base[buf->eop].upper.fields.status & E1000_TXD_STAT_DD : 0);
623 
624 	}
625 	}
626 #endif
627 
628         rc = sbuf_finish(sb);
629 	sbuf_delete(sb);
630         return(rc);
631 }
632 
633 static void *
634 em_register(device_t dev)
635 {
636 	return (em_sctx);
637 }
638 
639 static void *
640 igb_register(device_t dev)
641 {
642 	return (igb_sctx);
643 }
644 
645 static void
646 em_init_tx_ring(struct em_tx_queue *que)
647 {
648 	struct adapter *sc = que->adapter;
649 	if_softc_ctx_t scctx = sc->shared;
650 	struct tx_ring *txr = &que->txr;
651 	struct em_txbuffer *tx_buffer;
652 
653 	tx_buffer = txr->tx_buffers;
654 	for (int i = 0; i < scctx->isc_ntxd[0]; i++, tx_buffer++) {
655 		tx_buffer->eop = -1;
656 	}
657 }
658 
659 static int
660 em_set_num_queues(if_ctx_t ctx)
661 {
662 	struct adapter *adapter = iflib_get_softc(ctx);
663 	int maxqueues;
664 
665 	/* Sanity check based on HW */
666 	switch (adapter->hw.mac.type) {
667 		case e1000_82576:
668 		case e1000_82580:
669 		case e1000_i350:
670 		case e1000_i354:
671 			maxqueues = 8;
672 			break;
673 		case e1000_i210:
674 		case e1000_82575:
675 			maxqueues = 4;
676 			break;
677 		case e1000_i211:
678 		case e1000_82574:
679 			maxqueues = 2;
680 			break;
681 		default:
682 			maxqueues = 1;
683 			break;
684 	}
685 
686 	return (maxqueues);
687 }
688 
689 
690 #define EM_CAPS								\
691 	IFCAP_TSO4 | IFCAP_TXCSUM | IFCAP_LRO | IFCAP_RXCSUM | IFCAP_VLAN_HWFILTER | IFCAP_WOL_MAGIC | \
692 	IFCAP_WOL_MCAST | IFCAP_WOL | IFCAP_VLAN_HWTSO | IFCAP_HWCSUM | IFCAP_VLAN_HWTAGGING | \
693 	IFCAP_VLAN_HWCSUM | IFCAP_VLAN_HWTSO | IFCAP_VLAN_MTU;
694 
695 #define IGB_CAPS							\
696 	IFCAP_TSO4 | IFCAP_TXCSUM | IFCAP_LRO | IFCAP_RXCSUM | IFCAP_VLAN_HWFILTER | IFCAP_WOL_MAGIC | \
697 	IFCAP_WOL_MCAST | IFCAP_WOL | IFCAP_VLAN_HWTSO | IFCAP_HWCSUM | IFCAP_VLAN_HWTAGGING | IFCAP_VLAN_HWCSUM | \
698 	IFCAP_VLAN_HWTSO | IFCAP_VLAN_MTU | IFCAP_TXCSUM_IPV6 | IFCAP_HWCSUM_IPV6 | IFCAP_JUMBO_MTU;
699 
700 /*********************************************************************
701  *  Device initialization routine
702  *
703  *  The attach entry point is called when the driver is being loaded.
704  *  This routine identifies the type of hardware, allocates all resources
705  *  and initializes the hardware.
706  *
707  *  return 0 on success, positive on failure
708  *********************************************************************/
709 
710 static int
711 em_if_attach_pre(if_ctx_t ctx)
712 {
713         struct adapter	*adapter;
714 	if_softc_ctx_t scctx;
715         device_t        dev;
716 	struct e1000_hw	*hw;
717 	int		error = 0;
718 
719 	INIT_DEBUGOUT("em_if_attach_pre begin");
720         dev = iflib_get_dev(ctx);
721 	adapter = iflib_get_softc(ctx);
722 
723 	if (resource_disabled("em", device_get_unit(dev))) {
724 		device_printf(dev, "Disabled by device hint\n");
725 		return (ENXIO);
726 	}
727 
728 	adapter->ctx = ctx;
729 	adapter->dev = adapter->osdep.dev = dev;
730 	scctx = adapter->shared = iflib_get_softc_ctx(ctx);
731 	adapter->media = iflib_get_media(ctx);
732         hw = &adapter->hw;
733 
734 	adapter->tx_process_limit = scctx->isc_ntxd[0];
735 
736 	/* SYSCTL stuff */
737 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
738 	    SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
739 	    OID_AUTO, "nvm", CTLTYPE_INT|CTLFLAG_RW, adapter, 0,
740 	    em_sysctl_nvm_info, "I", "NVM Information");
741 
742 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
743 	    SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
744 	    OID_AUTO, "debug", CTLTYPE_INT|CTLFLAG_RW, adapter, 0,
745 	    em_sysctl_debug_info, "I", "Debug Information");
746 
747 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
748 	    SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
749 	    OID_AUTO, "fc", CTLTYPE_INT|CTLFLAG_RW, adapter, 0,
750 	    em_set_flowcntl, "I", "Flow Control");
751 
752 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
753 	    SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
754 	    OID_AUTO, "reg_dump", CTLTYPE_STRING | CTLFLAG_RD, adapter, 0,
755 	    em_get_regs, "A", "Dump Registers");
756 
757 	/* Determine hardware and mac info */
758 	em_identify_hardware(ctx);
759 
760         /* Set isc_msix_bar */
761 	scctx->isc_msix_bar = PCIR_BAR(EM_MSIX_BAR);
762 	scctx->isc_tx_nsegments = EM_MAX_SCATTER;
763 	scctx->isc_tx_tso_segments_max = scctx->isc_tx_nsegments;
764 	scctx->isc_tx_tso_size_max = EM_TSO_SIZE;
765 	scctx->isc_tx_tso_segsize_max = EM_TSO_SEG_SIZE;
766 	scctx->isc_nrxqsets_max = scctx->isc_ntxqsets_max = em_set_num_queues(ctx);
767 	device_printf(dev, "attach_pre capping queues at %d\n", scctx->isc_ntxqsets_max);
768 
769 	scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP | CSUM_IP_TSO;
770 
771 
772 	if (adapter->hw.mac.type >= igb_mac_min) {
773 		scctx->isc_txqsizes[0] = roundup2(scctx->isc_ntxd[0] * sizeof(union e1000_adv_tx_desc), EM_DBA_ALIGN);
774 		scctx->isc_rxqsizes[0] = roundup2(scctx->isc_nrxd[0] * sizeof(union e1000_adv_rx_desc), EM_DBA_ALIGN);
775 		scctx->isc_txrx = &igb_txrx;
776 		scctx->isc_capenable = IGB_CAPS;
777 		scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP | CSUM_TSO | CSUM_IP6_TCP \
778 			| CSUM_IP6_UDP | CSUM_IP6_TCP;
779 		if (adapter->hw.mac.type != e1000_82575)
780 			scctx->isc_tx_csum_flags |= CSUM_SCTP | CSUM_IP6_SCTP;
781 
782 	} else if (adapter->hw.mac.type >= em_mac_min) {
783 		scctx->isc_txqsizes[0] = roundup2(scctx->isc_ntxd[0]* sizeof(struct e1000_tx_desc), EM_DBA_ALIGN);
784 		scctx->isc_rxqsizes[0] = roundup2(scctx->isc_nrxd[0] * sizeof(union e1000_rx_desc_extended), EM_DBA_ALIGN);
785 		scctx->isc_txrx = &em_txrx;
786 		scctx->isc_capenable = EM_CAPS;
787 		scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP | CSUM_IP_TSO;
788 	} else {
789 		scctx->isc_txqsizes[0] = roundup2((scctx->isc_ntxd[0] + 1) * sizeof(struct e1000_tx_desc), EM_DBA_ALIGN);
790 		scctx->isc_rxqsizes[0] = roundup2((scctx->isc_nrxd[0] + 1) * sizeof(struct e1000_rx_desc), EM_DBA_ALIGN);
791 		scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP | CSUM_IP_TSO;
792 		scctx->isc_txrx = &lem_txrx;
793 		scctx->isc_capenable = EM_CAPS;
794 		if (adapter->hw.mac.type < e1000_82543)
795 			scctx->isc_capenable &= ~(IFCAP_HWCSUM|IFCAP_VLAN_HWCSUM);
796 		scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP | CSUM_IP_TSO;
797 		scctx->isc_msix_bar = 0;
798 	}
799 
800 	/* Setup PCI resources */
801 	if (em_allocate_pci_resources(ctx)) {
802 		device_printf(dev, "Allocation of PCI resources failed\n");
803 		error = ENXIO;
804 		goto err_pci;
805 	}
806 
807 	/*
808 	** For ICH8 and family we need to
809 	** map the flash memory, and this
810 	** must happen after the MAC is
811 	** identified
812 	*/
813 	if ((hw->mac.type == e1000_ich8lan) ||
814 	    (hw->mac.type == e1000_ich9lan) ||
815 	    (hw->mac.type == e1000_ich10lan) ||
816 	    (hw->mac.type == e1000_pchlan) ||
817 	    (hw->mac.type == e1000_pch2lan) ||
818 	    (hw->mac.type == e1000_pch_lpt)) {
819 		int rid = EM_BAR_TYPE_FLASH;
820 		adapter->flash = bus_alloc_resource_any(dev,
821 		    SYS_RES_MEMORY, &rid, RF_ACTIVE);
822 		if (adapter->flash == NULL) {
823 			device_printf(dev, "Mapping of Flash failed\n");
824 			error = ENXIO;
825 			goto err_pci;
826 		}
827 		/* This is used in the shared code */
828 		hw->flash_address = (u8 *)adapter->flash;
829 		adapter->osdep.flash_bus_space_tag =
830 		    rman_get_bustag(adapter->flash);
831 		adapter->osdep.flash_bus_space_handle =
832 		    rman_get_bushandle(adapter->flash);
833 	}
834 	/*
835 	** In the new SPT device flash is not  a
836 	** separate BAR, rather it is also in BAR0,
837 	** so use the same tag and an offset handle for the
838 	** FLASH read/write macros in the shared code.
839 	*/
840 	else if (hw->mac.type == e1000_pch_spt) {
841 		adapter->osdep.flash_bus_space_tag =
842 		    adapter->osdep.mem_bus_space_tag;
843 		adapter->osdep.flash_bus_space_handle =
844 		    adapter->osdep.mem_bus_space_handle
845 		    + E1000_FLASH_BASE_ADDR;
846 	}
847 
848 	/* Do Shared Code initialization */
849 	error = e1000_setup_init_funcs(hw, TRUE);
850 	if (error) {
851 		device_printf(dev, "Setup of Shared code failed, error %d\n",
852 		    error);
853 		error = ENXIO;
854 		goto err_pci;
855 	}
856 
857 	em_setup_msix(ctx);
858 	e1000_get_bus_info(hw);
859 
860 	/* Set up some sysctls for the tunable interrupt delays */
861 	em_add_int_delay_sysctl(adapter, "rx_int_delay",
862 	    "receive interrupt delay in usecs", &adapter->rx_int_delay,
863 	    E1000_REGISTER(hw, E1000_RDTR), em_rx_int_delay_dflt);
864 	em_add_int_delay_sysctl(adapter, "tx_int_delay",
865 	    "transmit interrupt delay in usecs", &adapter->tx_int_delay,
866 	    E1000_REGISTER(hw, E1000_TIDV), em_tx_int_delay_dflt);
867 	em_add_int_delay_sysctl(adapter, "rx_abs_int_delay",
868 	    "receive interrupt delay limit in usecs",
869 	    &adapter->rx_abs_int_delay,
870 	    E1000_REGISTER(hw, E1000_RADV),
871 	    em_rx_abs_int_delay_dflt);
872 	em_add_int_delay_sysctl(adapter, "tx_abs_int_delay",
873 	    "transmit interrupt delay limit in usecs",
874 	    &adapter->tx_abs_int_delay,
875 	    E1000_REGISTER(hw, E1000_TADV),
876 	    em_tx_abs_int_delay_dflt);
877 	em_add_int_delay_sysctl(adapter, "itr",
878 	    "interrupt delay limit in usecs/4",
879 	    &adapter->tx_itr,
880 	    E1000_REGISTER(hw, E1000_ITR),
881 	    DEFAULT_ITR);
882 
883 	/* Sysctl for limiting the amount of work done in the taskqueue */
884 	em_set_sysctl_value(adapter, "rx_processing_limit",
885 	    "max number of rx packets to process", &adapter->rx_process_limit,
886 	    em_rx_process_limit);
887 
888 	hw->mac.autoneg = DO_AUTO_NEG;
889 	hw->phy.autoneg_wait_to_complete = FALSE;
890 	hw->phy.autoneg_advertised = AUTONEG_ADV_DEFAULT;
891 
892 	if (adapter->hw.mac.type < em_mac_min) {
893 		e1000_init_script_state_82541(&adapter->hw, TRUE);
894 		e1000_set_tbi_compatibility_82543(&adapter->hw, TRUE);
895 	}
896 	/* Copper options */
897 	if (hw->phy.media_type == e1000_media_type_copper) {
898 		hw->phy.mdix = AUTO_ALL_MODES;
899 		hw->phy.disable_polarity_correction = FALSE;
900 		hw->phy.ms_type = EM_MASTER_SLAVE;
901 	}
902 
903 	/*
904 	 * Set the frame limits assuming
905 	 * standard ethernet sized frames.
906 	 */
907 	adapter->hw.mac.max_frame_size =
908 	    ETHERMTU + ETHER_HDR_LEN + ETHERNET_FCS_SIZE;
909 
910 	/*
911 	 * This controls when hardware reports transmit completion
912 	 * status.
913 	 */
914 	hw->mac.report_tx_early = 1;
915 
916 	/* Allocate multicast array memory. */
917 	adapter->mta = malloc(sizeof(u8) * ETH_ADDR_LEN *
918 	    MAX_NUM_MULTICAST_ADDRESSES, M_DEVBUF, M_NOWAIT);
919 	if (adapter->mta == NULL) {
920 		device_printf(dev, "Can not allocate multicast setup array\n");
921 		error = ENOMEM;
922 		goto err_late;
923 	}
924 
925 	/* Check SOL/IDER usage */
926 	if (e1000_check_reset_block(hw))
927 		device_printf(dev, "PHY reset is blocked"
928 		    " due to SOL/IDER session.\n");
929 
930 	/* Sysctl for setting Energy Efficient Ethernet */
931 	hw->dev_spec.ich8lan.eee_disable = eee_setting;
932 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
933 	    SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
934 	    OID_AUTO, "eee_control", CTLTYPE_INT|CTLFLAG_RW,
935 	    adapter, 0, em_sysctl_eee, "I",
936 	    "Disable Energy Efficient Ethernet");
937 
938 	/*
939 	** Start from a known state, this is
940 	** important in reading the nvm and
941 	** mac from that.
942 	*/
943 	e1000_reset_hw(hw);
944 
945 	/* Make sure we have a good EEPROM before we read from it */
946 	if (e1000_validate_nvm_checksum(hw) < 0) {
947 		/*
948 		** Some PCI-E parts fail the first check due to
949 		** the link being in sleep state, call it again,
950 		** if it fails a second time its a real issue.
951 		*/
952 		if (e1000_validate_nvm_checksum(hw) < 0) {
953 			device_printf(dev,
954 			    "The EEPROM Checksum Is Not Valid\n");
955 			error = EIO;
956 			goto err_late;
957 		}
958 	}
959 
960 	/* Copy the permanent MAC address out of the EEPROM */
961 	if (e1000_read_mac_addr(hw) < 0) {
962 		device_printf(dev, "EEPROM read error while reading MAC"
963 		    " address\n");
964 		error = EIO;
965 		goto err_late;
966 	}
967 
968 	if (!em_is_valid_ether_addr(hw->mac.addr)) {
969 		device_printf(dev, "Invalid MAC address\n");
970 		error = EIO;
971 		goto err_late;
972 	}
973 
974 	/* Disable ULP support */
975 	e1000_disable_ulp_lpt_lp(hw, TRUE);
976 
977         /*
978 	 * Get Wake-on-Lan and Management info for later use
979 	 */
980 	em_get_wakeup(ctx);
981 
982 	iflib_set_mac(ctx, hw->mac.addr);
983 
984 	return (0);
985 
986 err_late:
987 	em_release_hw_control(adapter);
988 err_pci:
989 	em_free_pci_resources(ctx);
990 	free(adapter->mta, M_DEVBUF);
991 
992 	return (error);
993 }
994 
995 static int
996 em_if_attach_post(if_ctx_t ctx)
997 {
998         struct adapter *adapter = iflib_get_softc(ctx);
999 	struct e1000_hw *hw = &adapter->hw;
1000  	int error = 0;
1001 
1002 	/* Setup OS specific network interface */
1003 	error = em_setup_interface(ctx);
1004 	if (error != 0) {
1005 		goto err_late;
1006 	}
1007 
1008 	em_reset(ctx);
1009 
1010 	/* Initialize statistics */
1011 	em_update_stats_counters(adapter);
1012 	hw->mac.get_link_status = 1;
1013 	em_if_update_admin_status(ctx);
1014 	em_add_hw_stats(adapter);
1015 
1016 	/* Non-AMT based hardware can now take control from firmware */
1017 	if (adapter->has_manage && !adapter->has_amt)
1018 		em_get_hw_control(adapter);
1019 
1020 	INIT_DEBUGOUT("em_if_attach_post: end");
1021 
1022 	return (error);
1023 
1024 err_late:
1025 	em_release_hw_control(adapter);
1026 	em_free_pci_resources(ctx);
1027 	em_if_queues_free(ctx);
1028 	free(adapter->mta, M_DEVBUF);
1029 
1030 	return (error);
1031 }
1032 
1033 /*********************************************************************
1034  *  Device removal routine
1035  *
1036  *  The detach entry point is called when the driver is being removed.
1037  *  This routine stops the adapter and deallocates all the resources
1038  *  that were allocated for driver operation.
1039  *
1040  *  return 0 on success, positive on failure
1041  *********************************************************************/
1042 
1043 static int
1044 em_if_detach(if_ctx_t ctx)
1045 {
1046 	struct adapter	*adapter = iflib_get_softc(ctx);
1047 
1048 	INIT_DEBUGOUT("em_detach: begin");
1049 
1050 	e1000_phy_hw_reset(&adapter->hw);
1051 
1052 	em_release_manageability(adapter);
1053 	em_release_hw_control(adapter);
1054 	em_free_pci_resources(ctx);
1055 
1056 	return (0);
1057 }
1058 
1059 /*********************************************************************
1060  *
1061  *  Shutdown entry point
1062  *
1063  **********************************************************************/
1064 
1065 static int
1066 em_if_shutdown(if_ctx_t ctx)
1067 {
1068 	return em_if_suspend(ctx);
1069 }
1070 
1071 /*
1072  * Suspend/resume device methods.
1073  */
1074 static int
1075 em_if_suspend(if_ctx_t ctx)
1076 {
1077 	struct adapter *adapter = iflib_get_softc(ctx);
1078 
1079         em_release_manageability(adapter);
1080 	em_release_hw_control(adapter);
1081 	em_enable_wakeup(ctx);
1082 	return (0);
1083 }
1084 
1085 static int
1086 em_if_resume(if_ctx_t ctx)
1087 {
1088 	struct adapter *adapter = iflib_get_softc(ctx);
1089 
1090 	if (adapter->hw.mac.type == e1000_pch2lan)
1091 		e1000_resume_workarounds_pchlan(&adapter->hw);
1092 	em_if_init(ctx);
1093 	em_init_manageability(adapter);
1094 
1095 	return(0);
1096 }
1097 
1098 static int
1099 em_if_mtu_set(if_ctx_t ctx, uint32_t mtu)
1100 {
1101   int max_frame_size;
1102   struct adapter *adapter = iflib_get_softc(ctx);
1103   struct ifnet *ifp = iflib_get_ifp(ctx);
1104 
1105   IOCTL_DEBUGOUT("ioctl rcv'd: SIOCSIFMTU (Set Interface MTU)");
1106 
1107   switch (adapter->hw.mac.type) {
1108   case e1000_82571:
1109   case e1000_82572:
1110   case e1000_ich9lan:
1111   case e1000_ich10lan:
1112   case e1000_pch2lan:
1113   case e1000_pch_lpt:
1114   case e1000_pch_spt:
1115   case e1000_82574:
1116   case e1000_82583:
1117   case e1000_80003es2lan:	/* 9K Jumbo Frame size */
1118 	  max_frame_size = 9234;
1119 	  break;
1120   case e1000_pchlan:
1121 	  max_frame_size = 4096;
1122 	  break;
1123 	  /* Adapters that do not support jumbo frames */
1124   case e1000_ich8lan:
1125 	  max_frame_size = ETHER_MAX_LEN;
1126 	  break;
1127   default:
1128 	  max_frame_size = MAX_JUMBO_FRAME_SIZE;
1129   }
1130   if (mtu > max_frame_size - ETHER_HDR_LEN - ETHER_CRC_LEN) {
1131 	  return (EINVAL);
1132   }
1133 
1134   adapter->hw.mac.max_frame_size = if_getmtu(ifp) + ETHER_HDR_LEN + ETHER_CRC_LEN;
1135   return (0);
1136 }
1137 
1138 /*********************************************************************
1139  *  Init entry point
1140  *
1141  *  This routine is used in two ways. It is used by the stack as
1142  *  init entry point in network interface structure. It is also used
1143  *  by the driver as a hw/sw initialization routine to get to a
1144  *  consistent state.
1145  *
1146  *  return 0 on success, positive on failure
1147  **********************************************************************/
1148 
1149 static void
1150 em_if_init(if_ctx_t ctx)
1151 {
1152         struct adapter *adapter = iflib_get_softc(ctx);
1153 	struct ifnet *ifp = iflib_get_ifp(ctx);
1154 
1155 	INIT_DEBUGOUT("em_if_init: begin");
1156 
1157 	/* Get the latest mac address, User can use a LAA */
1158         bcopy(if_getlladdr(ifp), adapter->hw.mac.addr,
1159               ETHER_ADDR_LEN);
1160 
1161 	/* Put the address into the Receive Address Array */
1162 	e1000_rar_set(&adapter->hw, adapter->hw.mac.addr, 0);
1163 
1164 	/*
1165 	 * With the 82571 adapter, RAR[0] may be overwritten
1166 	 * when the other port is reset, we make a duplicate
1167 	 * in RAR[14] for that eventuality, this assures
1168 	 * the interface continues to function.
1169 	 */
1170 	if (adapter->hw.mac.type == e1000_82571) {
1171 		e1000_set_laa_state_82571(&adapter->hw, TRUE);
1172 		e1000_rar_set(&adapter->hw, adapter->hw.mac.addr,
1173 		    E1000_RAR_ENTRIES - 1);
1174 	}
1175 
1176 	/* Initialize the hardware */
1177 	em_reset(ctx);
1178 	em_if_update_admin_status(ctx);
1179 
1180 	/* Setup VLAN support, basic and offload if available */
1181 	E1000_WRITE_REG(&adapter->hw, E1000_VET, ETHERTYPE_VLAN);
1182 
1183 	/* Clear bad data from Rx FIFOs */
1184 	if (adapter->hw.mac.type >= igb_mac_min)
1185 		e1000_rx_fifo_flush_82575(&adapter->hw);
1186 
1187 	/* Configure for OS presence */
1188 	em_init_manageability(adapter);
1189 
1190 	/* Prepare transmit descriptors and buffers */
1191 	em_initialize_transmit_unit(ctx);
1192 
1193 	/* Setup Multicast table */
1194 	em_if_multi_set(ctx);
1195 
1196 	/*
1197 	** Figure out the desired mbuf
1198 	** pool for doing jumbos
1199 	*/
1200 	if (adapter->hw.mac.max_frame_size <= 2048)
1201 		adapter->rx_mbuf_sz = MCLBYTES;
1202 	else if (adapter->hw.mac.max_frame_size <= 4096)
1203 		adapter->rx_mbuf_sz = MJUMPAGESIZE;
1204 	else
1205 		adapter->rx_mbuf_sz = MJUM9BYTES;
1206 
1207 	em_initialize_receive_unit(ctx);
1208 
1209 	/* Use real VLAN Filter support? */
1210 	if (if_getcapenable(ifp) & IFCAP_VLAN_HWTAGGING) {
1211 		if (if_getcapenable(ifp) & IFCAP_VLAN_HWFILTER)
1212 			/* Use real VLAN Filter support */
1213 			em_setup_vlan_hw_support(adapter);
1214 		else {
1215 			u32 ctrl;
1216 			ctrl = E1000_READ_REG(&adapter->hw, E1000_CTRL);
1217 			ctrl |= E1000_CTRL_VME;
1218 			E1000_WRITE_REG(&adapter->hw, E1000_CTRL, ctrl);
1219 		}
1220 	}
1221 
1222 	/* Don't lose promiscuous settings */
1223 	em_if_set_promisc(ctx, IFF_PROMISC);
1224 	e1000_clear_hw_cntrs_base_generic(&adapter->hw);
1225 
1226 	/* MSI/X configuration for 82574 */
1227 	if (adapter->hw.mac.type == e1000_82574) {
1228 		int tmp = E1000_READ_REG(&adapter->hw, E1000_CTRL_EXT);
1229 
1230 		tmp |= E1000_CTRL_EXT_PBA_CLR;
1231 		E1000_WRITE_REG(&adapter->hw, E1000_CTRL_EXT, tmp);
1232 		/* Set the IVAR - interrupt vector routing. */
1233 		E1000_WRITE_REG(&adapter->hw, E1000_IVAR, adapter->ivars);
1234 	} else if (adapter->intr_type == IFLIB_INTR_MSIX) /* Set up queue routing */
1235 		igb_configure_queues(adapter);
1236 
1237 	/* this clears any pending interrupts */
1238 	E1000_READ_REG(&adapter->hw, E1000_ICR);
1239 	E1000_WRITE_REG(&adapter->hw, E1000_ICS, E1000_ICS_LSC);
1240 
1241 	/* AMT based hardware can now take control from firmware */
1242 	if (adapter->has_manage && adapter->has_amt)
1243 		em_get_hw_control(adapter);
1244 
1245 	/* Set Energy Efficient Ethernet */
1246 	if (adapter->hw.mac.type >= igb_mac_min &&
1247 	    adapter->hw.phy.media_type == e1000_media_type_copper) {
1248 		if (adapter->hw.mac.type == e1000_i354)
1249 			e1000_set_eee_i354(&adapter->hw, TRUE, TRUE);
1250 		else
1251 			e1000_set_eee_i350(&adapter->hw, TRUE, TRUE);
1252 	}
1253 }
1254 
1255 /*********************************************************************
1256  *
1257  *  Fast Legacy/MSI Combined Interrupt Service routine
1258  *
1259  *********************************************************************/
1260 int
1261 em_intr(void *arg)
1262 {
1263         struct adapter	*adapter = arg;
1264 	if_ctx_t ctx = adapter->ctx;
1265 	u32		reg_icr;
1266 
1267 	reg_icr = E1000_READ_REG(&adapter->hw, E1000_ICR);
1268 
1269 	if (adapter->intr_type != IFLIB_INTR_LEGACY)
1270 		goto skip_stray;
1271 	/* Hot eject?  */
1272 	if (reg_icr == 0xffffffff)
1273 		return FILTER_STRAY;
1274 
1275 	/* Definitely not our interrupt.  */
1276 	if (reg_icr == 0x0)
1277 		return FILTER_STRAY;
1278 
1279 	/*
1280 	 * Starting with the 82571 chip, bit 31 should be used to
1281 	 * determine whether the interrupt belongs to us.
1282 	 */
1283 	if (adapter->hw.mac.type >= e1000_82571 &&
1284 	    (reg_icr & E1000_ICR_INT_ASSERTED) == 0)
1285 		return FILTER_STRAY;
1286 
1287 skip_stray:
1288 	/* Link status change */
1289 	if (reg_icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) {
1290 		adapter->hw.mac.get_link_status = 1;
1291 		iflib_admin_intr_deferred(ctx);
1292 	}
1293 
1294 	if (reg_icr & E1000_ICR_RXO)
1295 		adapter->rx_overruns++;
1296 
1297 	return (FILTER_SCHEDULE_THREAD);
1298 }
1299 
1300 static void
1301 igb_enable_queue(struct adapter *adapter, struct em_rx_queue *rxq)
1302 {
1303 	E1000_WRITE_REG(&adapter->hw, E1000_EIMS, rxq->eims);
1304 }
1305 
1306 static void
1307 em_enable_queue(struct adapter *adapter, struct em_rx_queue *rxq)
1308 {
1309 	E1000_WRITE_REG(&adapter->hw, E1000_IMS, rxq->eims);
1310 }
1311 
1312 static int
1313 em_if_queue_intr_enable(if_ctx_t ctx, uint16_t rxqid)
1314 {
1315         struct adapter	*adapter = iflib_get_softc(ctx);
1316 	struct em_rx_queue *rxq = &adapter->rx_queues[rxqid];
1317 
1318 	if (adapter->hw.mac.type >= igb_mac_min)
1319 		igb_enable_queue(adapter, rxq);
1320 	else
1321 		em_enable_queue(adapter, rxq);
1322 	return (0);
1323 }
1324 
1325 /*********************************************************************
1326  *
1327  *  MSIX RX Interrupt Service routine
1328  *
1329  **********************************************************************/
1330 static int
1331 em_msix_que(void *arg)
1332 {
1333 	struct em_rx_queue *que = arg;
1334 
1335 	++que->irqs;
1336 
1337         return (FILTER_SCHEDULE_THREAD);
1338 }
1339 
1340 /*********************************************************************
1341  *
1342  *  MSIX Link Fast Interrupt Service routine
1343  *
1344  **********************************************************************/
1345 static int
1346 em_msix_link(void *arg)
1347 {
1348 	struct adapter	*adapter = arg;
1349 	u32		reg_icr;
1350 
1351 	++adapter->link_irq;
1352 	MPASS(adapter->hw.back != NULL);
1353 	reg_icr = E1000_READ_REG(&adapter->hw, E1000_ICR);
1354 
1355 	if (reg_icr & E1000_ICR_RXO)
1356 		adapter->rx_overruns++;
1357 
1358 	if (reg_icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) {
1359 		em_handle_link(adapter->ctx);
1360 	} else {
1361 		E1000_WRITE_REG(&adapter->hw, E1000_IMS,
1362 				EM_MSIX_LINK | E1000_IMS_LSC);
1363 		if (adapter->hw.mac.type >= igb_mac_min)
1364 			E1000_WRITE_REG(&adapter->hw, E1000_EIMS, adapter->link_mask);
1365 
1366 	}
1367 
1368 	/*
1369  	** Because we must read the ICR for this interrupt
1370  	** it may clear other causes using autoclear, for
1371  	** this reason we simply create a soft interrupt
1372  	** for all these vectors.
1373  	*/
1374 	if (reg_icr && adapter->hw.mac.type < igb_mac_min) {
1375 		E1000_WRITE_REG(&adapter->hw,
1376 			E1000_ICS, adapter->ims);
1377 	}
1378 
1379 	return (FILTER_HANDLED);
1380 }
1381 
1382 static void
1383 em_handle_link(void *context)
1384 {
1385         if_ctx_t ctx = context;
1386 	struct adapter	*adapter = iflib_get_softc(ctx);
1387 
1388 	adapter->hw.mac.get_link_status = 1;
1389 	iflib_admin_intr_deferred(ctx);
1390 }
1391 
1392 
1393 /*********************************************************************
1394  *
1395  *  Media Ioctl callback
1396  *
1397  *  This routine is called whenever the user queries the status of
1398  *  the interface using ifconfig.
1399  *
1400  **********************************************************************/
1401 static void
1402 em_if_media_status(if_ctx_t ctx, struct ifmediareq *ifmr)
1403 {
1404   struct adapter *adapter = iflib_get_softc(ctx);
1405   u_char fiber_type = IFM_1000_SX;
1406 
1407   INIT_DEBUGOUT("em_if_media_status: begin");
1408 
1409         iflib_admin_intr_deferred(ctx);
1410 
1411 	ifmr->ifm_status = IFM_AVALID;
1412 	ifmr->ifm_active = IFM_ETHER;
1413 
1414 	if (!adapter->link_active) {
1415 		return;
1416 	}
1417 
1418 	ifmr->ifm_status |= IFM_ACTIVE;
1419 
1420 	if ((adapter->hw.phy.media_type == e1000_media_type_fiber) ||
1421 	    (adapter->hw.phy.media_type == e1000_media_type_internal_serdes)) {
1422 		if (adapter->hw.mac.type == e1000_82545)
1423 			fiber_type = IFM_1000_LX;
1424 		ifmr->ifm_active |= fiber_type | IFM_FDX;
1425 	} else {
1426 		switch (adapter->link_speed) {
1427 		case 10:
1428 			ifmr->ifm_active |= IFM_10_T;
1429 			break;
1430 		case 100:
1431 			ifmr->ifm_active |= IFM_100_TX;
1432 			break;
1433 		case 1000:
1434 			ifmr->ifm_active |= IFM_1000_T;
1435 			break;
1436 		}
1437 		if (adapter->link_duplex == FULL_DUPLEX)
1438 			ifmr->ifm_active |= IFM_FDX;
1439 		else
1440 			ifmr->ifm_active |= IFM_HDX;
1441 	}
1442 }
1443 
1444 /*********************************************************************
1445  *
1446  *  Media Ioctl callback
1447  *
1448  *  This routine is called when the user changes speed/duplex using
1449  *  media/mediopt option with ifconfig.
1450  *
1451  **********************************************************************/
1452 static int
1453 em_if_media_change(if_ctx_t ctx)
1454 {
1455 	struct adapter *adapter = iflib_get_softc(ctx);
1456 	struct ifmedia  *ifm = iflib_get_media(ctx);
1457 
1458 	INIT_DEBUGOUT("em_if_media_change: begin");
1459 
1460 	if (IFM_TYPE(ifm->ifm_media) != IFM_ETHER)
1461 		return (EINVAL);
1462 
1463 	switch (IFM_SUBTYPE(ifm->ifm_media)) {
1464 	case IFM_AUTO:
1465 		adapter->hw.mac.autoneg = DO_AUTO_NEG;
1466 		adapter->hw.phy.autoneg_advertised = AUTONEG_ADV_DEFAULT;
1467 		break;
1468 	case IFM_1000_LX:
1469 	case IFM_1000_SX:
1470 	case IFM_1000_T:
1471 		adapter->hw.mac.autoneg = DO_AUTO_NEG;
1472 		adapter->hw.phy.autoneg_advertised = ADVERTISE_1000_FULL;
1473 		break;
1474 	case IFM_100_TX:
1475 		adapter->hw.mac.autoneg = FALSE;
1476 		adapter->hw.phy.autoneg_advertised = 0;
1477 		if ((ifm->ifm_media & IFM_GMASK) == IFM_FDX)
1478 			adapter->hw.mac.forced_speed_duplex = ADVERTISE_100_FULL;
1479 		else
1480 			adapter->hw.mac.forced_speed_duplex = ADVERTISE_100_HALF;
1481 		break;
1482 	case IFM_10_T:
1483 		adapter->hw.mac.autoneg = FALSE;
1484 		adapter->hw.phy.autoneg_advertised = 0;
1485 		if ((ifm->ifm_media & IFM_GMASK) == IFM_FDX)
1486 			adapter->hw.mac.forced_speed_duplex = ADVERTISE_10_FULL;
1487 		else
1488 			adapter->hw.mac.forced_speed_duplex = ADVERTISE_10_HALF;
1489 		break;
1490 	default:
1491 		device_printf(adapter->dev, "Unsupported media type\n");
1492 	}
1493 
1494 	em_if_init(ctx);
1495 
1496 	return (0);
1497 }
1498 
1499 static int
1500 em_if_set_promisc(if_ctx_t ctx, int flags)
1501 {
1502         struct adapter *adapter = iflib_get_softc(ctx);
1503 	u32		reg_rctl;
1504 
1505 	em_disable_promisc(ctx);
1506 
1507 	reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
1508 
1509 	if (flags & IFF_PROMISC) {
1510 		reg_rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
1511 		/* Turn this on if you want to see bad packets */
1512 		if (em_debug_sbp)
1513 			reg_rctl |= E1000_RCTL_SBP;
1514 		E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1515 	} else if (flags & IFF_ALLMULTI) {
1516 		reg_rctl |= E1000_RCTL_MPE;
1517 		reg_rctl &= ~E1000_RCTL_UPE;
1518 		E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1519 	}
1520 	return (0);
1521 }
1522 
1523 static void
1524 em_disable_promisc(if_ctx_t ctx)
1525 {
1526         struct adapter *adapter = iflib_get_softc(ctx);
1527 	struct ifnet *ifp = iflib_get_ifp(ctx);
1528 	u32		reg_rctl;
1529 	int		mcnt = 0;
1530 
1531 	reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
1532 	reg_rctl &=  (~E1000_RCTL_UPE);
1533 	if (if_getflags(ifp) & IFF_ALLMULTI)
1534 		mcnt = MAX_NUM_MULTICAST_ADDRESSES;
1535 	else
1536 		mcnt = if_multiaddr_count(ifp, MAX_NUM_MULTICAST_ADDRESSES);
1537 	/* Don't disable if in MAX groups */
1538 	if (mcnt < MAX_NUM_MULTICAST_ADDRESSES)
1539 		reg_rctl &=  (~E1000_RCTL_MPE);
1540 	reg_rctl &=  (~E1000_RCTL_SBP);
1541 	E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1542 }
1543 
1544 
1545 /*********************************************************************
1546  *  Multicast Update
1547  *
1548  *  This routine is called whenever multicast address list is updated.
1549  *
1550  **********************************************************************/
1551 
1552 static void
1553 em_if_multi_set(if_ctx_t ctx)
1554 {
1555         struct adapter *adapter = iflib_get_softc(ctx);
1556 	struct ifnet *ifp = iflib_get_ifp(ctx);
1557 	u32 reg_rctl = 0;
1558 	u8  *mta; /* Multicast array memory */
1559 	int mcnt = 0;
1560 
1561 	IOCTL_DEBUGOUT("em_set_multi: begin");
1562 
1563 	mta = adapter->mta;
1564 	bzero(mta, sizeof(u8) * ETH_ADDR_LEN * MAX_NUM_MULTICAST_ADDRESSES);
1565 
1566 	if (adapter->hw.mac.type == e1000_82542 &&
1567 	    adapter->hw.revision_id == E1000_REVISION_2) {
1568 		reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
1569 		if (adapter->hw.bus.pci_cmd_word & CMD_MEM_WRT_INVALIDATE)
1570 			e1000_pci_clear_mwi(&adapter->hw);
1571 		reg_rctl |= E1000_RCTL_RST;
1572 		E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1573 		msec_delay(5);
1574 	}
1575 
1576 	if_multiaddr_array(ifp, mta, &mcnt, MAX_NUM_MULTICAST_ADDRESSES);
1577 
1578 	if (mcnt >= MAX_NUM_MULTICAST_ADDRESSES) {
1579 		reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
1580 		reg_rctl |= E1000_RCTL_MPE;
1581 		E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1582 	} else
1583 		e1000_update_mc_addr_list(&adapter->hw, mta, mcnt);
1584 
1585 	if (adapter->hw.mac.type == e1000_82542 &&
1586 	    adapter->hw.revision_id == E1000_REVISION_2) {
1587 		reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
1588 		reg_rctl &= ~E1000_RCTL_RST;
1589 		E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1590 		msec_delay(5);
1591 		if (adapter->hw.bus.pci_cmd_word & CMD_MEM_WRT_INVALIDATE)
1592 			e1000_pci_set_mwi(&adapter->hw);
1593 	}
1594 }
1595 
1596 
1597 /*********************************************************************
1598  *  Timer routine
1599  *
1600  *  This routine checks for link status and updates statistics.
1601  *
1602  **********************************************************************/
1603 
1604 static void
1605 em_if_timer(if_ctx_t ctx, uint16_t qid)
1606 {
1607         struct adapter	*adapter = iflib_get_softc(ctx);
1608 	struct em_rx_queue *que;
1609 	int i;
1610 	int trigger = 0;
1611 
1612 	em_if_update_admin_status(ctx);
1613 	em_update_stats_counters(adapter);
1614 
1615 	/* Reset LAA into RAR[0] on 82571 */
1616 	if ((adapter->hw.mac.type == e1000_82571) &&
1617 	    e1000_get_laa_state_82571(&adapter->hw))
1618 		e1000_rar_set(&adapter->hw, adapter->hw.mac.addr, 0);
1619 
1620 	if (adapter->hw.mac.type < em_mac_min)
1621 		lem_smartspeed(adapter);
1622 
1623 	/* Mask to use in the irq trigger */
1624 	if (adapter->intr_type == IFLIB_INTR_MSIX) {
1625 		for (i = 0, que = adapter->rx_queues; i < adapter->rx_num_queues; i++, que++)
1626 			trigger |= que->eims;
1627 	} else {
1628 		trigger = E1000_ICS_RXDMT0;
1629 	}
1630 }
1631 
1632 
1633 static void
1634 em_if_update_admin_status(if_ctx_t ctx)
1635 {
1636         struct adapter *adapter = iflib_get_softc(ctx);
1637 	struct e1000_hw *hw = &adapter->hw;
1638         struct ifnet *ifp = iflib_get_ifp(ctx);
1639 	device_t dev = iflib_get_dev(ctx);
1640 	u32 link_check = 0;
1641 
1642 	/* Get the cached link value or read phy for real */
1643 	switch (hw->phy.media_type) {
1644 	case e1000_media_type_copper:
1645 		if (hw->mac.get_link_status) {
1646 			if (hw->mac.type == e1000_pch_spt)
1647 				msec_delay(50);
1648 			/* Do the work to read phy */
1649 			e1000_check_for_link(hw);
1650 			link_check = !hw->mac.get_link_status;
1651 			if (link_check) /* ESB2 fix */
1652 				e1000_cfg_on_link_up(hw);
1653 		} else {
1654 			link_check = TRUE;
1655 		}
1656 		break;
1657 	case e1000_media_type_fiber:
1658 		e1000_check_for_link(hw);
1659 		link_check = (E1000_READ_REG(hw, E1000_STATUS) &
1660                                  E1000_STATUS_LU);
1661 		break;
1662 	case e1000_media_type_internal_serdes:
1663 		e1000_check_for_link(hw);
1664 		link_check = adapter->hw.mac.serdes_has_link;
1665 		break;
1666 	default:
1667 	case e1000_media_type_unknown:
1668 		break;
1669 	}
1670 
1671 	/* Now check for a transition */
1672 	if (link_check && (adapter->link_active == 0)) {
1673 		e1000_get_speed_and_duplex(hw, &adapter->link_speed,
1674 		    &adapter->link_duplex);
1675 		/* Check if we must disable SPEED_MODE bit on PCI-E */
1676 		if ((adapter->link_speed != SPEED_1000) &&
1677 		    ((hw->mac.type == e1000_82571) ||
1678 		    (hw->mac.type == e1000_82572))) {
1679 			int tarc0;
1680 			tarc0 = E1000_READ_REG(hw, E1000_TARC(0));
1681 			tarc0 &= ~TARC_SPEED_MODE_BIT;
1682 			E1000_WRITE_REG(hw, E1000_TARC(0), tarc0);
1683 		}
1684 		if (bootverbose)
1685 			device_printf(dev, "Link is up %d Mbps %s\n",
1686 			    adapter->link_speed,
1687 			    ((adapter->link_duplex == FULL_DUPLEX) ?
1688 			    "Full Duplex" : "Half Duplex"));
1689 		adapter->link_active = 1;
1690 		adapter->smartspeed = 0;
1691 		if_setbaudrate(ifp, adapter->link_speed * 1000000);
1692 		iflib_link_state_change(ctx, LINK_STATE_UP, ifp->if_baudrate);
1693 		printf("Link state changed to up\n");
1694 	} else if (!link_check && (adapter->link_active == 1)) {
1695 		if_setbaudrate(ifp, 0);
1696 		adapter->link_speed = 0;
1697 		adapter->link_duplex = 0;
1698 		if (bootverbose)
1699 			device_printf(dev, "Link is Down\n");
1700 		adapter->link_active = 0;
1701 		iflib_link_state_change(ctx, LINK_STATE_DOWN, ifp->if_baudrate);
1702 		printf("link state changed to down\n");
1703 	}
1704 
1705 	E1000_WRITE_REG(&adapter->hw, E1000_IMS, EM_MSIX_LINK | E1000_IMS_LSC);
1706 }
1707 
1708 /*********************************************************************
1709  *
1710  *  This routine disables all traffic on the adapter by issuing a
1711  *  global reset on the MAC and deallocates TX/RX buffers.
1712  *
1713  *  This routine should always be called with BOTH the CORE
1714  *  and TX locks.
1715  **********************************************************************/
1716 
1717 static void
1718 em_if_stop(if_ctx_t ctx)
1719 {
1720         struct adapter *adapter = iflib_get_softc(ctx);
1721 
1722 	INIT_DEBUGOUT("em_stop: begin");
1723 
1724 	e1000_reset_hw(&adapter->hw);
1725 	if (adapter->hw.mac.type >= e1000_82544)
1726 		E1000_WRITE_REG(&adapter->hw, E1000_WUC, 0);
1727 
1728 	e1000_led_off(&adapter->hw);
1729 	e1000_cleanup_led(&adapter->hw);
1730 }
1731 
1732 
1733 /*********************************************************************
1734  *
1735  *  Determine hardware revision.
1736  *
1737  **********************************************************************/
1738 static void
1739 em_identify_hardware(if_ctx_t ctx)
1740 {
1741         device_t dev = iflib_get_dev(ctx);
1742         struct adapter *adapter = iflib_get_softc(ctx);
1743 
1744 	/* Make sure our PCI config space has the necessary stuff set */
1745 	adapter->hw.bus.pci_cmd_word = pci_read_config(dev, PCIR_COMMAND, 2);
1746 
1747 	/* Save off the information about this board */
1748 	adapter->hw.vendor_id = pci_get_vendor(dev);
1749 	adapter->hw.device_id = pci_get_device(dev);
1750 	adapter->hw.revision_id = pci_read_config(dev, PCIR_REVID, 1);
1751 	adapter->hw.subsystem_vendor_id =
1752 	    pci_read_config(dev, PCIR_SUBVEND_0, 2);
1753 	adapter->hw.subsystem_device_id =
1754 	    pci_read_config(dev, PCIR_SUBDEV_0, 2);
1755 
1756 	/* Do Shared Code Init and Setup */
1757 	if (e1000_set_mac_type(&adapter->hw)) {
1758 		device_printf(dev, "Setup init failure\n");
1759 		return;
1760 	}
1761 }
1762 
1763 static int
1764 em_allocate_pci_resources(if_ctx_t ctx)
1765 {
1766         struct adapter *adapter = iflib_get_softc(ctx);
1767         device_t	dev = iflib_get_dev(ctx);
1768 	int		rid, val;
1769 
1770 	rid = PCIR_BAR(0);
1771 	adapter->memory = bus_alloc_resource_any(dev, SYS_RES_MEMORY,
1772 	    &rid, RF_ACTIVE);
1773 	if (adapter->memory == NULL) {
1774 		device_printf(dev, "Unable to allocate bus resource: memory\n");
1775 		return (ENXIO);
1776 	}
1777 	adapter->osdep.mem_bus_space_tag =
1778 	    rman_get_bustag(adapter->memory);
1779 	adapter->osdep.mem_bus_space_handle =
1780 	    rman_get_bushandle(adapter->memory);
1781 	adapter->hw.hw_addr = (u8 *)&adapter->osdep.mem_bus_space_handle;
1782 
1783 	/* Only older adapters use IO mapping */
1784 	if (adapter->hw.mac.type < em_mac_min &&
1785 	    adapter->hw.mac.type > e1000_82543) {
1786 		/* Figure our where our IO BAR is ? */
1787 		for (rid = PCIR_BAR(0); rid < PCIR_CIS;) {
1788 			val = pci_read_config(dev, rid, 4);
1789 			if (EM_BAR_TYPE(val) == EM_BAR_TYPE_IO) {
1790 				adapter->io_rid = rid;
1791 				break;
1792 			}
1793 			rid += 4;
1794 			/* check for 64bit BAR */
1795 			if (EM_BAR_MEM_TYPE(val) == EM_BAR_MEM_TYPE_64BIT)
1796 				rid += 4;
1797 		}
1798 		if (rid >= PCIR_CIS) {
1799 			device_printf(dev, "Unable to locate IO BAR\n");
1800 			return (ENXIO);
1801 		}
1802 		adapter->ioport = bus_alloc_resource_any(dev,
1803 		    SYS_RES_IOPORT, &adapter->io_rid, RF_ACTIVE);
1804 		if (adapter->ioport == NULL) {
1805 			device_printf(dev, "Unable to allocate bus resource: "
1806 			    "ioport\n");
1807 			return (ENXIO);
1808 		}
1809 		adapter->hw.io_base = 0;
1810 		adapter->osdep.io_bus_space_tag =
1811 		    rman_get_bustag(adapter->ioport);
1812 		adapter->osdep.io_bus_space_handle =
1813 		    rman_get_bushandle(adapter->ioport);
1814 	}
1815 
1816 	adapter->hw.back = &adapter->osdep;
1817 
1818 	return (0);
1819 }
1820 
1821 /*********************************************************************
1822  *
1823  *  Setup the MSIX Interrupt handlers
1824  *
1825  **********************************************************************/
1826 static int
1827 em_if_msix_intr_assign(if_ctx_t ctx, int msix)
1828 {
1829         struct adapter     *adapter = iflib_get_softc(ctx);
1830 	struct em_rx_queue *rx_que = adapter->rx_queues;
1831 	struct em_tx_queue *tx_que = adapter->tx_queues;
1832 	int		   error, rid, i, vector = 0;
1833 	char buf[16];
1834 
1835 	/* First set up ring resources */
1836 	for (i = 0; i < adapter->rx_num_queues; i++, rx_que++, vector++) {
1837 		rid = vector +1;
1838 		snprintf(buf, sizeof(buf), "rxq%d", i);
1839 		error = iflib_irq_alloc_generic(ctx, &rx_que->que_irq, rid, IFLIB_INTR_RX, em_msix_que, rx_que, rx_que->me, buf);
1840                 if (error) {
1841 		  	device_printf(iflib_get_dev(ctx), "Failed to allocate que int %d err: %d", i, error);
1842 			adapter->rx_num_queues = i + 1;
1843 			goto fail;
1844 		}
1845 
1846 		rx_que->msix =  vector;
1847 
1848 		/*
1849 		** Set the bit to enable interrupt
1850 		** in E1000_IMS -- bits 20 and 21
1851 		** are for RX0 and RX1, note this has
1852 		** NOTHING to do with the MSIX vector
1853 		*/
1854 		if (adapter->hw.mac.type == e1000_82574) {
1855 			rx_que->eims = 1 << (20 + i);
1856 			adapter->ims |= rx_que->eims;
1857 			adapter->ivars |= (8 | rx_que->msix) << (i * 4);
1858 		} else if (adapter->hw.mac.type == e1000_82575)
1859 			rx_que->eims = E1000_EICR_TX_QUEUE0 << vector;
1860 		else
1861 			rx_que->eims = 1 << vector;
1862 	}
1863 
1864 	for (i = 0; i < adapter->tx_num_queues; i++, tx_que++) {
1865 		rid = vector + 1;
1866 		snprintf(buf, sizeof(buf), "txq%d", i);
1867 		tx_que = &adapter->tx_queues[i];
1868 		iflib_softirq_alloc_generic(ctx, rid, IFLIB_INTR_TX, tx_que, tx_que->me, buf);
1869 
1870 		tx_que->msix = vector;
1871 
1872 		  /*
1873 		** Set the bit to enable interrupt
1874 		** in E1000_IMS -- bits 22 and 23
1875 		** are for TX0 and TX1, note this has
1876 		** NOTHING to do with the MSIX vector
1877 		*/
1878 		if (adapter->hw.mac.type < igb_mac_min) {
1879 			tx_que->eims = 1 << (22 + i);
1880 			adapter->ims |= tx_que->eims;
1881 			adapter->ivars |= (8 | tx_que->msix) << (8 + (i * 4));
1882 		} if (adapter->hw.mac.type == e1000_82575)
1883 			tx_que->eims = E1000_EICR_TX_QUEUE0 << (i %  adapter->tx_num_queues);
1884 		else
1885 			tx_que->eims = 1 << (i %  adapter->tx_num_queues);
1886 	}
1887 
1888 	/* Link interrupt */
1889 	rid = vector + 1;
1890         error = iflib_irq_alloc_generic(ctx, &adapter->irq, rid, IFLIB_INTR_ADMIN, em_msix_link, adapter, 0, "aq");
1891 
1892 	if (error) {
1893 		device_printf(iflib_get_dev(ctx), "Failed to register admin handler");
1894 		goto fail;
1895 	}
1896 	adapter->linkvec = vector;
1897 	if (adapter->hw.mac.type < igb_mac_min) {
1898 		adapter->ivars |=  (8 | vector) << 16;
1899 		adapter->ivars |= 0x80000000;
1900 	}
1901 	return (0);
1902  fail:
1903 	iflib_irq_free(ctx, &adapter->irq);
1904 	rx_que = adapter->rx_queues;
1905 	for (int i = 0; i < adapter->rx_num_queues; i++, rx_que++)
1906 		iflib_irq_free(ctx, &rx_que->que_irq);
1907 	return (error);
1908 }
1909 
1910 static void
1911 igb_configure_queues(struct adapter *adapter)
1912 {
1913 	struct	e1000_hw	*hw = &adapter->hw;
1914 	struct	em_rx_queue	*rx_que;
1915 	struct  em_tx_queue    *tx_que;
1916 	u32			tmp, ivar = 0, newitr = 0;
1917 
1918 	/* First turn on RSS capability */
1919 	if (adapter->hw.mac.type != e1000_82575)
1920 		E1000_WRITE_REG(hw, E1000_GPIE,
1921 		    E1000_GPIE_MSIX_MODE | E1000_GPIE_EIAME |
1922 		    E1000_GPIE_PBA | E1000_GPIE_NSICR);
1923 
1924 	/* Turn on MSIX */
1925 	switch (adapter->hw.mac.type) {
1926 	case e1000_82580:
1927 	case e1000_i350:
1928 	case e1000_i354:
1929 	case e1000_i210:
1930 	case e1000_i211:
1931 	case e1000_vfadapt:
1932 	case e1000_vfadapt_i350:
1933 		/* RX entries */
1934 		for (int i = 0; i < adapter->rx_num_queues; i++) {
1935 			u32 index = i >> 1;
1936 			ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
1937 			rx_que = &adapter->rx_queues[i];
1938 			if (i & 1) {
1939 				ivar &= 0xFF00FFFF;
1940 				ivar |= (rx_que->msix | E1000_IVAR_VALID) << 16;
1941 			} else {
1942 				ivar &= 0xFFFFFF00;
1943 				ivar |= rx_que->msix | E1000_IVAR_VALID;
1944 			}
1945 			E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
1946 		}
1947 		/* TX entries */
1948 		for (int i = 0; i < adapter->tx_num_queues; i++) {
1949 			u32 index = i >> 1;
1950 			ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
1951 			tx_que = &adapter->tx_queues[i];
1952 			if (i & 1) {
1953 				ivar &= 0x00FFFFFF;
1954 				ivar |= (tx_que->msix | E1000_IVAR_VALID) << 24;
1955 			} else {
1956 				ivar &= 0xFFFF00FF;
1957 				ivar |= (tx_que->msix | E1000_IVAR_VALID) << 8;
1958 			}
1959 			E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
1960 			adapter->que_mask |= tx_que->eims;
1961 		}
1962 
1963 		/* And for the link interrupt */
1964 		ivar = (adapter->linkvec | E1000_IVAR_VALID) << 8;
1965 		adapter->link_mask = 1 << adapter->linkvec;
1966 		E1000_WRITE_REG(hw, E1000_IVAR_MISC, ivar);
1967 		break;
1968 	case e1000_82576:
1969 		/* RX entries */
1970 		for (int i = 0; i < adapter->rx_num_queues; i++) {
1971 			u32 index = i & 0x7; /* Each IVAR has two entries */
1972 			ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
1973 			rx_que = &adapter->rx_queues[i];
1974 			if (i < 8) {
1975 				ivar &= 0xFFFFFF00;
1976 				ivar |= rx_que->msix | E1000_IVAR_VALID;
1977 			} else {
1978 				ivar &= 0xFF00FFFF;
1979 				ivar |= (rx_que->msix | E1000_IVAR_VALID) << 16;
1980 			}
1981 			E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
1982 			adapter->que_mask |= rx_que->eims;
1983 		}
1984 		/* TX entries */
1985 		for (int i = 0; i < adapter->tx_num_queues; i++) {
1986 			u32 index = i & 0x7; /* Each IVAR has two entries */
1987 			ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
1988 			tx_que = &adapter->tx_queues[i];
1989 			if (i < 8) {
1990 				ivar &= 0xFFFF00FF;
1991 				ivar |= (tx_que->msix | E1000_IVAR_VALID) << 8;
1992 			} else {
1993 				ivar &= 0x00FFFFFF;
1994 				ivar |= (tx_que->msix | E1000_IVAR_VALID) << 24;
1995 			}
1996 			E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
1997 			adapter->que_mask |= tx_que->eims;
1998 		}
1999 
2000 		/* And for the link interrupt */
2001 		ivar = (adapter->linkvec | E1000_IVAR_VALID) << 8;
2002 		adapter->link_mask = 1 << adapter->linkvec;
2003 		E1000_WRITE_REG(hw, E1000_IVAR_MISC, ivar);
2004 		break;
2005 
2006 	case e1000_82575:
2007                 /* enable MSI-X support*/
2008 		tmp = E1000_READ_REG(hw, E1000_CTRL_EXT);
2009                 tmp |= E1000_CTRL_EXT_PBA_CLR;
2010                 /* Auto-Mask interrupts upon ICR read. */
2011                 tmp |= E1000_CTRL_EXT_EIAME;
2012                 tmp |= E1000_CTRL_EXT_IRCA;
2013                 E1000_WRITE_REG(hw, E1000_CTRL_EXT, tmp);
2014 
2015 		/* Queues */
2016 		for (int i = 0; i < adapter->rx_num_queues; i++) {
2017 			rx_que = &adapter->rx_queues[i];
2018 			tmp = E1000_EICR_RX_QUEUE0 << i;
2019 			tmp |= E1000_EICR_TX_QUEUE0 << i;
2020 			rx_que->eims = tmp;
2021 			E1000_WRITE_REG_ARRAY(hw, E1000_MSIXBM(0),
2022 			    i, rx_que->eims);
2023 			adapter->que_mask |= rx_que->eims;
2024 		}
2025 
2026 		/* Link */
2027 		E1000_WRITE_REG(hw, E1000_MSIXBM(adapter->linkvec),
2028 		    E1000_EIMS_OTHER);
2029 		adapter->link_mask |= E1000_EIMS_OTHER;
2030 	default:
2031 		break;
2032 	}
2033 
2034 	/* Set the starting interrupt rate */
2035 	if (em_max_interrupt_rate > 0)
2036 		newitr = (4000000 / em_max_interrupt_rate) & 0x7FFC;
2037 
2038         if (hw->mac.type == e1000_82575)
2039                 newitr |= newitr << 16;
2040         else
2041                 newitr |= E1000_EITR_CNT_IGNR;
2042 
2043 	for (int i = 0; i < adapter->rx_num_queues; i++) {
2044 		rx_que = &adapter->rx_queues[i];
2045 		E1000_WRITE_REG(hw, E1000_EITR(rx_que->msix), newitr);
2046 	}
2047 
2048 	return;
2049 }
2050 
2051 static void
2052 em_free_pci_resources(if_ctx_t ctx)
2053 {
2054 	struct adapter *adapter = iflib_get_softc(ctx);
2055         struct 		em_rx_queue *que = adapter->rx_queues;
2056 	device_t	dev = iflib_get_dev(ctx);
2057 
2058 	/* Release all msix queue resources */
2059 	if (adapter->intr_type == IFLIB_INTR_MSIX)
2060 		iflib_irq_free(ctx, &adapter->irq);
2061 
2062 	for (int i = 0; i < adapter->rx_num_queues; i++, que++) {
2063 		iflib_irq_free(ctx, &que->que_irq);
2064 	}
2065 
2066 
2067 	/* First release all the interrupt resources */
2068 	if (adapter->memory != NULL) {
2069 		bus_release_resource(dev, SYS_RES_MEMORY,
2070 				     PCIR_BAR(0), adapter->memory);
2071 		adapter->memory = NULL;
2072 	}
2073 
2074 	if (adapter->flash != NULL) {
2075 		bus_release_resource(dev, SYS_RES_MEMORY,
2076 				     EM_FLASH, adapter->flash);
2077 		adapter->flash = NULL;
2078 	}
2079 	if (adapter->ioport != NULL)
2080 		bus_release_resource(dev, SYS_RES_IOPORT,
2081 		    adapter->io_rid, adapter->ioport);
2082 }
2083 
2084 /* Setup MSI or MSI/X */
2085 static int
2086 em_setup_msix(if_ctx_t ctx)
2087 {
2088 	struct adapter *adapter = iflib_get_softc(ctx);
2089 
2090 	if (adapter->hw.mac.type == e1000_82574) {
2091 		em_enable_vectors_82574(ctx);
2092 	}
2093 	return (0);
2094 }
2095 
2096 /*********************************************************************
2097  *
2098  *  Initialize the hardware to a configuration
2099  *  as specified by the adapter structure.
2100  *
2101  **********************************************************************/
2102 
2103 static void
2104 lem_smartspeed(struct adapter *adapter)
2105 {
2106 	u16 phy_tmp;
2107 
2108 	if (adapter->link_active || (adapter->hw.phy.type != e1000_phy_igp) ||
2109 	    adapter->hw.mac.autoneg == 0 ||
2110 	    (adapter->hw.phy.autoneg_advertised & ADVERTISE_1000_FULL) == 0)
2111 		return;
2112 
2113 	if (adapter->smartspeed == 0) {
2114 		/* If Master/Slave config fault is asserted twice,
2115 		 * we assume back-to-back */
2116 		e1000_read_phy_reg(&adapter->hw, PHY_1000T_STATUS, &phy_tmp);
2117 		if (!(phy_tmp & SR_1000T_MS_CONFIG_FAULT))
2118 			return;
2119 		e1000_read_phy_reg(&adapter->hw, PHY_1000T_STATUS, &phy_tmp);
2120 		if (phy_tmp & SR_1000T_MS_CONFIG_FAULT) {
2121 			e1000_read_phy_reg(&adapter->hw,
2122 			    PHY_1000T_CTRL, &phy_tmp);
2123 			if(phy_tmp & CR_1000T_MS_ENABLE) {
2124 				phy_tmp &= ~CR_1000T_MS_ENABLE;
2125 				e1000_write_phy_reg(&adapter->hw,
2126 				    PHY_1000T_CTRL, phy_tmp);
2127 				adapter->smartspeed++;
2128 				if(adapter->hw.mac.autoneg &&
2129 				   !e1000_copper_link_autoneg(&adapter->hw) &&
2130 				   !e1000_read_phy_reg(&adapter->hw,
2131 				    PHY_CONTROL, &phy_tmp)) {
2132 					phy_tmp |= (MII_CR_AUTO_NEG_EN |
2133 						    MII_CR_RESTART_AUTO_NEG);
2134 					e1000_write_phy_reg(&adapter->hw,
2135 					    PHY_CONTROL, phy_tmp);
2136 				}
2137 			}
2138 		}
2139 		return;
2140 	} else if(adapter->smartspeed == EM_SMARTSPEED_DOWNSHIFT) {
2141 		/* If still no link, perhaps using 2/3 pair cable */
2142 		e1000_read_phy_reg(&adapter->hw, PHY_1000T_CTRL, &phy_tmp);
2143 		phy_tmp |= CR_1000T_MS_ENABLE;
2144 		e1000_write_phy_reg(&adapter->hw, PHY_1000T_CTRL, phy_tmp);
2145 		if(adapter->hw.mac.autoneg &&
2146 		   !e1000_copper_link_autoneg(&adapter->hw) &&
2147 		   !e1000_read_phy_reg(&adapter->hw, PHY_CONTROL, &phy_tmp)) {
2148 			phy_tmp |= (MII_CR_AUTO_NEG_EN |
2149 				    MII_CR_RESTART_AUTO_NEG);
2150 			e1000_write_phy_reg(&adapter->hw, PHY_CONTROL, phy_tmp);
2151 		}
2152 	}
2153 	/* Restart process after EM_SMARTSPEED_MAX iterations */
2154 	if(adapter->smartspeed++ == EM_SMARTSPEED_MAX)
2155 		adapter->smartspeed = 0;
2156 }
2157 
2158 
2159 static void
2160 em_reset(if_ctx_t ctx)
2161 {
2162         device_t	dev = iflib_get_dev(ctx);
2163 	struct adapter *adapter = iflib_get_softc(ctx);
2164 	struct ifnet *ifp = iflib_get_ifp(ctx);
2165 	struct e1000_hw	*hw = &adapter->hw;
2166 	u16		rx_buffer_size;
2167 	u32		pba;
2168 
2169 	INIT_DEBUGOUT("em_reset: begin");
2170 
2171 	/* Set up smart power down as default off on newer adapters. */
2172 	if (!em_smart_pwr_down && (hw->mac.type == e1000_82571 ||
2173 	    hw->mac.type == e1000_82572)) {
2174 		u16 phy_tmp = 0;
2175 
2176 		/* Speed up time to link by disabling smart power down. */
2177 		e1000_read_phy_reg(hw, IGP02E1000_PHY_POWER_MGMT, &phy_tmp);
2178 		phy_tmp &= ~IGP02E1000_PM_SPD;
2179 		e1000_write_phy_reg(hw, IGP02E1000_PHY_POWER_MGMT, phy_tmp);
2180 	}
2181 
2182 	/*
2183 	 * Packet Buffer Allocation (PBA)
2184 	 * Writing PBA sets the receive portion of the buffer
2185 	 * the remainder is used for the transmit buffer.
2186 	 */
2187 	switch (hw->mac.type) {
2188 	/* Total Packet Buffer on these is 48K */
2189 	case e1000_82571:
2190 	case e1000_82572:
2191 	case e1000_80003es2lan:
2192 			pba = E1000_PBA_32K; /* 32K for Rx, 16K for Tx */
2193 		break;
2194 	case e1000_82573: /* 82573: Total Packet Buffer is 32K */
2195 			pba = E1000_PBA_12K; /* 12K for Rx, 20K for Tx */
2196 		break;
2197 	case e1000_82574:
2198 	case e1000_82583:
2199 			pba = E1000_PBA_20K; /* 20K for Rx, 20K for Tx */
2200 		break;
2201 	case e1000_ich8lan:
2202 		pba = E1000_PBA_8K;
2203 		break;
2204 	case e1000_ich9lan:
2205 	case e1000_ich10lan:
2206 		/* Boost Receive side for jumbo frames */
2207 		if (adapter->hw.mac.max_frame_size > 4096)
2208 			pba = E1000_PBA_14K;
2209 		else
2210 			pba = E1000_PBA_10K;
2211 		break;
2212 	case e1000_pchlan:
2213 	case e1000_pch2lan:
2214 	case e1000_pch_lpt:
2215 	case e1000_pch_spt:
2216 		pba = E1000_PBA_26K;
2217 		break;
2218 	default:
2219 		if (adapter->hw.mac.max_frame_size > 8192)
2220 			pba = E1000_PBA_40K; /* 40K for Rx, 24K for Tx */
2221 		else
2222 			pba = E1000_PBA_48K; /* 48K for Rx, 16K for Tx */
2223 	}
2224 	E1000_WRITE_REG(&adapter->hw, E1000_PBA, pba);
2225 
2226 	/*
2227 	 * These parameters control the automatic generation (Tx) and
2228 	 * response (Rx) to Ethernet PAUSE frames.
2229 	 * - High water mark should allow for at least two frames to be
2230 	 *   received after sending an XOFF.
2231 	 * - Low water mark works best when it is very near the high water mark.
2232 	 *   This allows the receiver to restart by sending XON when it has
2233 	 *   drained a bit. Here we use an arbitrary value of 1500 which will
2234 	 *   restart after one full frame is pulled from the buffer. There
2235 	 *   could be several smaller frames in the buffer and if so they will
2236 	 *   not trigger the XON until their total number reduces the buffer
2237 	 *   by 1500.
2238 	 * - The pause time is fairly large at 1000 x 512ns = 512 usec.
2239 	 */
2240 	rx_buffer_size = ((E1000_READ_REG(hw, E1000_PBA) & 0xffff) << 10 );
2241 	hw->fc.high_water = rx_buffer_size -
2242 	    roundup2(adapter->hw.mac.max_frame_size, 1024);
2243 	hw->fc.low_water = hw->fc.high_water - 1500;
2244 
2245 	if (adapter->fc) /* locally set flow control value? */
2246 		hw->fc.requested_mode = adapter->fc;
2247 	else
2248 		hw->fc.requested_mode = e1000_fc_full;
2249 
2250 	if (hw->mac.type == e1000_80003es2lan)
2251 		hw->fc.pause_time = 0xFFFF;
2252 	else
2253 		hw->fc.pause_time = EM_FC_PAUSE_TIME;
2254 
2255 	hw->fc.send_xon = TRUE;
2256 
2257 	/* Device specific overrides/settings */
2258 	switch (hw->mac.type) {
2259 	case e1000_pchlan:
2260 		/* Workaround: no TX flow ctrl for PCH */
2261                 hw->fc.requested_mode = e1000_fc_rx_pause;
2262 		hw->fc.pause_time = 0xFFFF; /* override */
2263 		if (if_getmtu(ifp) > ETHERMTU) {
2264 			hw->fc.high_water = 0x3500;
2265 			hw->fc.low_water = 0x1500;
2266 		} else {
2267 			hw->fc.high_water = 0x5000;
2268 			hw->fc.low_water = 0x3000;
2269 		}
2270 		hw->fc.refresh_time = 0x1000;
2271 		break;
2272 	case e1000_pch2lan:
2273 	case e1000_pch_lpt:
2274 	case e1000_pch_spt:
2275 		hw->fc.high_water = 0x5C20;
2276 		hw->fc.low_water = 0x5048;
2277 		hw->fc.pause_time = 0x0650;
2278 		hw->fc.refresh_time = 0x0400;
2279 		/* Jumbos need adjusted PBA */
2280 		if (if_getmtu(ifp) > ETHERMTU)
2281 			E1000_WRITE_REG(hw, E1000_PBA, 12);
2282 		else
2283 			E1000_WRITE_REG(hw, E1000_PBA, 26);
2284 		break;
2285         case e1000_ich9lan:
2286         case e1000_ich10lan:
2287 		if (if_getmtu(ifp) > ETHERMTU) {
2288 			hw->fc.high_water = 0x2800;
2289 			hw->fc.low_water = hw->fc.high_water - 8;
2290 			break;
2291 		}
2292 		/* else fall thru */
2293 	default:
2294 		if (hw->mac.type == e1000_80003es2lan)
2295 			hw->fc.pause_time = 0xFFFF;
2296 		break;
2297 	}
2298 
2299 	/* Issue a global reset */
2300 	e1000_reset_hw(hw);
2301 	E1000_WRITE_REG(hw, E1000_WUC, 0);
2302 	em_disable_aspm(adapter);
2303 	/* and a re-init */
2304 	if (e1000_init_hw(hw) < 0) {
2305 		device_printf(dev, "Hardware Initialization Failed\n");
2306 		return;
2307 	}
2308 
2309 	E1000_WRITE_REG(hw, E1000_VET, ETHERTYPE_VLAN);
2310 	e1000_get_phy_info(hw);
2311 	e1000_check_for_link(hw);
2312 }
2313 
2314 #define RSSKEYLEN 10
2315 static void
2316 em_initialize_rss_mapping(struct adapter *adapter)
2317 {
2318 	uint8_t  rss_key[4 * RSSKEYLEN];
2319 	uint32_t reta = 0;
2320 	struct e1000_hw	*hw = &adapter->hw;
2321 	int i;
2322 
2323 	/*
2324 	 * Configure RSS key
2325 	 */
2326 	arc4rand(rss_key, sizeof(rss_key), 0);
2327 	for (i = 0; i < RSSKEYLEN; ++i) {
2328 		uint32_t rssrk = 0;
2329 
2330 		rssrk = EM_RSSRK_VAL(rss_key, i);
2331 		E1000_WRITE_REG(hw,E1000_RSSRK(i), rssrk);
2332 	}
2333 
2334 	/*
2335 	 * Configure RSS redirect table in following fashion:
2336 	 * (hash & ring_cnt_mask) == rdr_table[(hash & rdr_table_mask)]
2337 	 */
2338 	for (i = 0; i < sizeof(reta); ++i) {
2339 		uint32_t q;
2340 
2341 		q = (i % adapter->rx_num_queues) << 7;
2342 		reta |= q << (8 * i);
2343 	}
2344 
2345 	for (i = 0; i < 32; ++i)
2346 		E1000_WRITE_REG(hw, E1000_RETA(i), reta);
2347 
2348 	E1000_WRITE_REG(hw, E1000_MRQC, E1000_MRQC_RSS_ENABLE_2Q |
2349 			E1000_MRQC_RSS_FIELD_IPV4_TCP |
2350 			E1000_MRQC_RSS_FIELD_IPV4 |
2351 			E1000_MRQC_RSS_FIELD_IPV6_TCP_EX |
2352 			E1000_MRQC_RSS_FIELD_IPV6_EX |
2353 			E1000_MRQC_RSS_FIELD_IPV6);
2354 
2355 }
2356 
2357 static void
2358 igb_initialize_rss_mapping(struct adapter *adapter)
2359 {
2360 	struct e1000_hw *hw = &adapter->hw;
2361 	int i;
2362 	int queue_id;
2363 	u32 reta;
2364 	u32 rss_key[10], mrqc, shift = 0;
2365 
2366 	/* XXX? */
2367 	if (adapter->hw.mac.type == e1000_82575)
2368 		shift = 6;
2369 
2370 	/*
2371 	 * The redirection table controls which destination
2372 	 * queue each bucket redirects traffic to.
2373 	 * Each DWORD represents four queues, with the LSB
2374 	 * being the first queue in the DWORD.
2375 	 *
2376 	 * This just allocates buckets to queues using round-robin
2377 	 * allocation.
2378 	 *
2379 	 * NOTE: It Just Happens to line up with the default
2380 	 * RSS allocation method.
2381 	 */
2382 
2383 	/* Warning FM follows */
2384 	reta = 0;
2385 	for (i = 0; i < 128; i++) {
2386 #ifdef	RSS
2387 		queue_id = rss_get_indirection_to_bucket(i);
2388 		/*
2389 		 * If we have more queues than buckets, we'll
2390 		 * end up mapping buckets to a subset of the
2391 		 * queues.
2392 		 *
2393 		 * If we have more buckets than queues, we'll
2394 		 * end up instead assigning multiple buckets
2395 		 * to queues.
2396 		 *
2397 		 * Both are suboptimal, but we need to handle
2398 		 * the case so we don't go out of bounds
2399 		 * indexing arrays and such.
2400 		 */
2401 		queue_id = queue_id % adapter->rx_num_queues;
2402 #else
2403 		queue_id = (i % adapter->rx_num_queues);
2404 #endif
2405 		/* Adjust if required */
2406 		queue_id = queue_id << shift;
2407 
2408 		/*
2409 		 * The low 8 bits are for hash value (n+0);
2410 		 * The next 8 bits are for hash value (n+1), etc.
2411 		 */
2412 		reta = reta >> 8;
2413 		reta = reta | ( ((uint32_t) queue_id) << 24);
2414 		if ((i & 3) == 3) {
2415 			E1000_WRITE_REG(hw, E1000_RETA(i >> 2), reta);
2416 			reta = 0;
2417 		}
2418 	}
2419 
2420 	/* Now fill in hash table */
2421 
2422 	/*
2423 	 * MRQC: Multiple Receive Queues Command
2424 	 * Set queuing to RSS control, number depends on the device.
2425 	 */
2426 	mrqc = E1000_MRQC_ENABLE_RSS_8Q;
2427 
2428 #ifdef	RSS
2429 	/* XXX ew typecasting */
2430 	rss_getkey((uint8_t *) &rss_key);
2431 #else
2432 	arc4rand(&rss_key, sizeof(rss_key), 0);
2433 #endif
2434 	for (i = 0; i < 10; i++)
2435 		E1000_WRITE_REG_ARRAY(hw,
2436 		    E1000_RSSRK(0), i, rss_key[i]);
2437 
2438 	/*
2439 	 * Configure the RSS fields to hash upon.
2440 	 */
2441 	mrqc |= (E1000_MRQC_RSS_FIELD_IPV4 |
2442 	    E1000_MRQC_RSS_FIELD_IPV4_TCP);
2443 	mrqc |= (E1000_MRQC_RSS_FIELD_IPV6 |
2444 	    E1000_MRQC_RSS_FIELD_IPV6_TCP);
2445 	mrqc |=( E1000_MRQC_RSS_FIELD_IPV4_UDP |
2446 	    E1000_MRQC_RSS_FIELD_IPV6_UDP);
2447 	mrqc |=( E1000_MRQC_RSS_FIELD_IPV6_UDP_EX |
2448 	    E1000_MRQC_RSS_FIELD_IPV6_TCP_EX);
2449 
2450 	E1000_WRITE_REG(hw, E1000_MRQC, mrqc);
2451 }
2452 
2453 /*********************************************************************
2454  *
2455  *  Setup networking device structure and register an interface.
2456  *
2457  **********************************************************************/
2458 static int
2459 em_setup_interface(if_ctx_t ctx)
2460 {
2461         struct ifnet *ifp = iflib_get_ifp(ctx);
2462         struct adapter *adapter = iflib_get_softc(ctx);
2463 	if_softc_ctx_t scctx = adapter->shared;
2464 	uint64_t cap = 0;
2465 
2466 	INIT_DEBUGOUT("em_setup_interface: begin");
2467 
2468 	/* TSO parameters */
2469 	ifp->if_hw_tsomax = IP_MAXPACKET;
2470 	/* Take m_pullup(9)'s in em_xmit() w/ TSO into acount. */
2471 	ifp->if_hw_tsomaxsegcount = EM_MAX_SCATTER - 5;
2472 	ifp->if_hw_tsomaxsegsize = EM_TSO_SEG_SIZE;
2473 
2474 	/* Single Queue */
2475         if (adapter->tx_num_queues == 1) {
2476 	  if_setsendqlen(ifp, scctx->isc_ntxd[0] - 1);
2477 	  if_setsendqready(ifp);
2478 	}
2479 
2480         cap = IFCAP_HWCSUM | IFCAP_VLAN_HWCSUM | IFCAP_TSO4;
2481 	cap |= IFCAP_VLAN_HWTAGGING | IFCAP_VLAN_HWTSO | IFCAP_VLAN_MTU;
2482 
2483 	/*
2484 	 * Tell the upper layer(s) we
2485 	 * support full VLAN capability
2486 	 */
2487 	if_setifheaderlen(ifp, sizeof(struct ether_vlan_header));
2488 	if_setcapabilitiesbit(ifp, cap, 0);
2489 
2490 	/*
2491 	** Don't turn this on by default, if vlans are
2492 	** created on another pseudo device (eg. lagg)
2493 	** then vlan events are not passed thru, breaking
2494 	** operation, but with HW FILTER off it works. If
2495 	** using vlans directly on the em driver you can
2496 	** enable this and get full hardware tag filtering.
2497 	*/
2498 	if_setcapabilitiesbit(ifp, IFCAP_VLAN_HWFILTER,0);
2499 
2500 	/* Enable only WOL MAGIC by default */
2501 	if (adapter->wol) {
2502 		if_setcapabilitiesbit(ifp, IFCAP_WOL, 0);
2503 		if_setcapenablebit(ifp, IFCAP_WOL_MAGIC, 0);
2504 	}
2505 
2506 	/*
2507 	 * Specify the media types supported by this adapter and register
2508 	 * callbacks to update media and link information
2509 	 */
2510 	if ((adapter->hw.phy.media_type == e1000_media_type_fiber) ||
2511 	    (adapter->hw.phy.media_type == e1000_media_type_internal_serdes)) {
2512 		u_char fiber_type = IFM_1000_SX;	/* default type */
2513 
2514 		if (adapter->hw.mac.type == e1000_82545)
2515 			fiber_type = IFM_1000_LX;
2516 		ifmedia_add(adapter->media, IFM_ETHER | fiber_type | IFM_FDX, 0, NULL);
2517 		ifmedia_add(adapter->media, IFM_ETHER | fiber_type, 0, NULL);
2518 	} else {
2519 		ifmedia_add(adapter->media, IFM_ETHER | IFM_10_T, 0, NULL);
2520 		ifmedia_add(adapter->media, IFM_ETHER | IFM_10_T | IFM_FDX, 0, NULL);
2521 		ifmedia_add(adapter->media, IFM_ETHER | IFM_100_TX, 0, NULL);
2522 		ifmedia_add(adapter->media, IFM_ETHER | IFM_100_TX | IFM_FDX, 0, NULL);
2523 		if (adapter->hw.phy.type != e1000_phy_ife) {
2524 			ifmedia_add(adapter->media, IFM_ETHER | IFM_1000_T | IFM_FDX, 0, NULL);
2525 			ifmedia_add(adapter->media, IFM_ETHER | IFM_1000_T, 0, NULL);
2526 		}
2527 	}
2528 	ifmedia_add(adapter->media, IFM_ETHER | IFM_AUTO, 0, NULL);
2529 	ifmedia_set(adapter->media, IFM_ETHER | IFM_AUTO);
2530 	return (0);
2531 }
2532 
2533 static int
2534 em_if_tx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int ntxqs, int ntxqsets)
2535 {
2536 	struct adapter *adapter = iflib_get_softc(ctx);
2537 	if_softc_ctx_t scctx = adapter->shared;
2538 	int error = E1000_SUCCESS;
2539 	struct em_tx_queue *que;
2540         int i;
2541 
2542 	MPASS(adapter->tx_num_queues > 0);
2543 	MPASS(adapter->tx_num_queues == ntxqsets);
2544 
2545 	/* First allocate the top level queue structs */
2546 	if (!(adapter->tx_queues =
2547 	    (struct em_tx_queue *) malloc(sizeof(struct em_tx_queue) *
2548 	    adapter->tx_num_queues, M_DEVBUF, M_NOWAIT | M_ZERO))) {
2549 		device_printf(iflib_get_dev(ctx), "Unable to allocate queue memory\n");
2550 		return(ENOMEM);
2551 	}
2552 
2553 	for (i = 0, que = adapter->tx_queues; i < adapter->tx_num_queues; i++, que++) {
2554 	     /* Set up some basics */
2555 	     struct tx_ring *txr = &que->txr;
2556 	     txr->adapter = que->adapter = adapter;
2557 	     txr->que = que;
2558 	     que->me = txr->me =  i;
2559 
2560 	     /* Allocate transmit buffer memory */
2561 	  if (!(txr->tx_buffers = (struct em_txbuffer *) malloc(sizeof(struct em_txbuffer) * scctx->isc_ntxd[0], M_DEVBUF, M_NOWAIT | M_ZERO))) {
2562 	       device_printf(iflib_get_dev(ctx), "failed to allocate tx_buffer memory\n");
2563 	       error = ENOMEM;
2564 	       goto fail;
2565 	  }
2566 
2567 	  /* get the virtual and physical address of the hardware queues */
2568 	  txr->tx_base = (struct e1000_tx_desc *)vaddrs[i*ntxqs];
2569 	  txr->tx_paddr = paddrs[i*ntxqs];
2570 
2571 	}
2572 
2573 	device_printf(iflib_get_dev(ctx), "allocated for %d tx_queues\n", adapter->tx_num_queues);
2574 	return (0);
2575  fail:
2576 	em_if_queues_free(ctx);
2577 	return (error);
2578 }
2579 
2580 static int
2581 em_if_rx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int nrxqs, int nrxqsets)
2582 {
2583         struct adapter *adapter = iflib_get_softc(ctx);
2584 	int error = E1000_SUCCESS;
2585 	struct em_rx_queue *que;
2586         int i;
2587 
2588 	MPASS(adapter->rx_num_queues > 0);
2589 	MPASS(adapter->rx_num_queues == nrxqsets);
2590 
2591 	/* First allocate the top level queue structs */
2592 	if (!(adapter->rx_queues =
2593 	    (struct em_rx_queue *) malloc(sizeof(struct em_rx_queue) *
2594 	    adapter->rx_num_queues, M_DEVBUF, M_NOWAIT | M_ZERO))) {
2595 		device_printf(iflib_get_dev(ctx), "Unable to allocate queue memory\n");
2596 		error = ENOMEM;
2597 		goto fail;
2598 	}
2599 
2600 	for (i = 0, que = adapter->rx_queues; i < nrxqsets; i++, que++) {
2601 		/* Set up some basics */
2602 		struct rx_ring *rxr = &que->rxr;
2603 		rxr->adapter = que->adapter = adapter;
2604 		rxr->que = que;
2605 		que->me = rxr->me =  i;
2606 
2607 		/* get the virtual and physical address of the hardware queues */
2608 		rxr->rx_base = (union e1000_rx_desc_extended *)vaddrs[i*nrxqs];
2609 		rxr->rx_paddr = paddrs[i*nrxqs];
2610 	}
2611 
2612 	device_printf(iflib_get_dev(ctx), "allocated for %d rx_queues\n", adapter->rx_num_queues);
2613 
2614 	return (0);
2615 fail:
2616 	em_if_queues_free(ctx);
2617 	return (error);
2618 }
2619 
2620 static void
2621 em_if_queues_free(if_ctx_t ctx)
2622 {
2623     	struct adapter *adapter = iflib_get_softc(ctx);
2624 	struct em_tx_queue *tx_que = adapter->tx_queues;
2625 	struct em_rx_queue *rx_que = adapter->rx_queues;
2626 
2627 	if (tx_que != NULL) {
2628 	  for (int i = 0; i < adapter->tx_num_queues; i++, tx_que++) {
2629 		struct tx_ring *txr = &tx_que->txr;
2630 		if (txr->tx_buffers == NULL)
2631 			break;
2632 
2633 		free(txr->tx_buffers, M_DEVBUF);
2634 		txr->tx_buffers = NULL;
2635 	  }
2636 	  free(adapter->tx_queues, M_DEVBUF);
2637 	  adapter->tx_queues = NULL;
2638 	}
2639 
2640 	if (rx_que != NULL) {
2641 	  free(adapter->rx_queues, M_DEVBUF);
2642 	  adapter->rx_queues = NULL;
2643 	}
2644 
2645 	em_release_hw_control(adapter);
2646 
2647 	if (adapter->mta != NULL) {
2648 		free(adapter->mta, M_DEVBUF);
2649 	}
2650 }
2651 
2652 /*********************************************************************
2653  *
2654  *  Enable transmit unit.
2655  *
2656  **********************************************************************/
2657 static void
2658 em_initialize_transmit_unit(if_ctx_t ctx)
2659 {
2660         struct adapter *adapter = iflib_get_softc(ctx);
2661 	if_softc_ctx_t scctx = adapter->shared;
2662 	struct em_tx_queue *que;
2663 	struct tx_ring	*txr;
2664 	struct e1000_hw	*hw = &adapter->hw;
2665 	u32 tctl, txdctl = 0, tarc, tipg = 0;
2666 
2667 	 INIT_DEBUGOUT("em_initialize_transmit_unit: begin");
2668 
2669 	for (int i = 0; i < adapter->tx_num_queues; i++, txr++) {
2670 		u64 bus_addr;
2671 		caddr_t offp, endp;
2672 
2673 	        que = &adapter->tx_queues[i];
2674 		txr = &que->txr;
2675 		bus_addr = txr->tx_paddr;
2676 
2677                 /*Enable all queues */
2678 		em_init_tx_ring(que);
2679 
2680 		/* Clear checksum offload context. */
2681 		offp = (caddr_t)&txr->csum_flags;
2682 		endp = (caddr_t)(txr + 1);
2683 		bzero(offp, endp - offp);
2684 
2685 		/* Base and Len of TX Ring */
2686 		E1000_WRITE_REG(hw, E1000_TDLEN(i),
2687 		    scctx->isc_ntxd[0] * sizeof(struct e1000_tx_desc));
2688 		E1000_WRITE_REG(hw, E1000_TDBAH(i),
2689 	    	    (u32)(bus_addr >> 32));
2690 		E1000_WRITE_REG(hw, E1000_TDBAL(i),
2691 	    	    (u32)bus_addr);
2692 		/* Init the HEAD/TAIL indices */
2693 		E1000_WRITE_REG(hw, E1000_TDT(i), 0);
2694 		E1000_WRITE_REG(hw, E1000_TDH(i), 0);
2695 
2696 		HW_DEBUGOUT2("Base = %x, Length = %x\n",
2697 		    E1000_READ_REG(&adapter->hw, E1000_TDBAL(i)),
2698 		    E1000_READ_REG(&adapter->hw, E1000_TDLEN(i)));
2699 
2700 		txdctl = 0; /* clear txdctl */
2701                 txdctl |= 0x1f; /* PTHRESH */
2702                 txdctl |= 1 << 8; /* HTHRESH */
2703                 txdctl |= 1 << 16;/* WTHRESH */
2704 		txdctl |= 1 << 22; /* Reserved bit 22 must always be 1 */
2705 		txdctl |= E1000_TXDCTL_GRAN;
2706                 txdctl |= 1 << 25; /* LWTHRESH */
2707 
2708                 E1000_WRITE_REG(hw, E1000_TXDCTL(i), txdctl);
2709 	}
2710 
2711 	/* Set the default values for the Tx Inter Packet Gap timer */
2712 	switch (adapter->hw.mac.type) {
2713 	case e1000_80003es2lan:
2714 		tipg = DEFAULT_82543_TIPG_IPGR1;
2715 		tipg |= DEFAULT_80003ES2LAN_TIPG_IPGR2 <<
2716 		    E1000_TIPG_IPGR2_SHIFT;
2717 		break;
2718 	case e1000_82542:
2719 		tipg = DEFAULT_82542_TIPG_IPGT;
2720 		tipg |= DEFAULT_82542_TIPG_IPGR1 << E1000_TIPG_IPGR1_SHIFT;
2721 		tipg |= DEFAULT_82542_TIPG_IPGR2 << E1000_TIPG_IPGR2_SHIFT;
2722 		break;
2723 	default:
2724 		if ((adapter->hw.phy.media_type == e1000_media_type_fiber) ||
2725 		    (adapter->hw.phy.media_type ==
2726 		    e1000_media_type_internal_serdes))
2727 			tipg = DEFAULT_82543_TIPG_IPGT_FIBER;
2728 		else
2729 			tipg = DEFAULT_82543_TIPG_IPGT_COPPER;
2730 		tipg |= DEFAULT_82543_TIPG_IPGR1 << E1000_TIPG_IPGR1_SHIFT;
2731 		tipg |= DEFAULT_82543_TIPG_IPGR2 << E1000_TIPG_IPGR2_SHIFT;
2732 	}
2733 
2734 	E1000_WRITE_REG(&adapter->hw, E1000_TIPG, tipg);
2735 	E1000_WRITE_REG(&adapter->hw, E1000_TIDV, adapter->tx_int_delay.value);
2736 
2737 	if(adapter->hw.mac.type >= e1000_82540)
2738 		E1000_WRITE_REG(&adapter->hw, E1000_TADV,
2739 		    adapter->tx_abs_int_delay.value);
2740 
2741 	if ((adapter->hw.mac.type == e1000_82571) ||
2742 	    (adapter->hw.mac.type == e1000_82572)) {
2743 		tarc = E1000_READ_REG(&adapter->hw, E1000_TARC(0));
2744 		tarc |= TARC_SPEED_MODE_BIT;
2745 		E1000_WRITE_REG(&adapter->hw, E1000_TARC(0), tarc);
2746 	} else if (adapter->hw.mac.type == e1000_80003es2lan) {
2747 		/* errata: program both queues to unweighted RR */
2748 		tarc = E1000_READ_REG(&adapter->hw, E1000_TARC(0));
2749 		tarc |= 1;
2750 		E1000_WRITE_REG(&adapter->hw, E1000_TARC(0), tarc);
2751 		tarc = E1000_READ_REG(&adapter->hw, E1000_TARC(1));
2752 		tarc |= 1;
2753 		E1000_WRITE_REG(&adapter->hw, E1000_TARC(1), tarc);
2754 	} else if (adapter->hw.mac.type == e1000_82574) {
2755 		tarc = E1000_READ_REG(&adapter->hw, E1000_TARC(0));
2756 		tarc |= TARC_ERRATA_BIT;
2757 		if ( adapter->tx_num_queues > 1) {
2758 			tarc |= (TARC_COMPENSATION_MODE | TARC_MQ_FIX);
2759 			E1000_WRITE_REG(&adapter->hw, E1000_TARC(0), tarc);
2760 			E1000_WRITE_REG(&adapter->hw, E1000_TARC(1), tarc);
2761 		} else
2762 			E1000_WRITE_REG(&adapter->hw, E1000_TARC(0), tarc);
2763 	}
2764 
2765 	if (adapter->tx_int_delay.value > 0)
2766 		adapter->txd_cmd |= E1000_TXD_CMD_IDE;
2767 
2768 	/* Program the Transmit Control Register */
2769 	tctl = E1000_READ_REG(&adapter->hw, E1000_TCTL);
2770 	tctl &= ~E1000_TCTL_CT;
2771 	tctl |= (E1000_TCTL_PSP | E1000_TCTL_RTLC | E1000_TCTL_EN |
2772 		   (E1000_COLLISION_THRESHOLD << E1000_CT_SHIFT));
2773 
2774 	if (adapter->hw.mac.type >= e1000_82571)
2775 		tctl |= E1000_TCTL_MULR;
2776 
2777 	/* This write will effectively turn on the transmit unit. */
2778 	E1000_WRITE_REG(&adapter->hw, E1000_TCTL, tctl);
2779 
2780 	if (hw->mac.type == e1000_pch_spt) {
2781 		u32 reg;
2782 		reg = E1000_READ_REG(hw, E1000_IOSFPC);
2783 		reg |= E1000_RCTL_RDMTS_HEX;
2784 		E1000_WRITE_REG(hw, E1000_IOSFPC, reg);
2785 		reg = E1000_READ_REG(hw, E1000_TARC(0));
2786 		reg |= E1000_TARC0_CB_MULTIQ_3_REQ;
2787 		E1000_WRITE_REG(hw, E1000_TARC(0), reg);
2788 	}
2789 }
2790 
2791 /*********************************************************************
2792  *
2793  *  Enable receive unit.
2794  *
2795  **********************************************************************/
2796 
2797 static void
2798 em_initialize_receive_unit(if_ctx_t ctx)
2799 {
2800         struct adapter *adapter = iflib_get_softc(ctx);
2801 	if_softc_ctx_t scctx = adapter->shared;
2802 	struct ifnet *ifp = iflib_get_ifp(ctx);
2803 	struct e1000_hw	*hw = &adapter->hw;
2804 	struct em_rx_queue *que;
2805 	int i;
2806 	u32	rctl, rxcsum, rfctl;
2807 
2808 	INIT_DEBUGOUT("em_initialize_receive_units: begin");
2809 
2810 	/*
2811 	 * Make sure receives are disabled while setting
2812 	 * up the descriptor ring
2813 	 */
2814 	rctl = E1000_READ_REG(hw, E1000_RCTL);
2815 	/* Do not disable if ever enabled on this hardware */
2816 	if ((hw->mac.type != e1000_82574) && (hw->mac.type != e1000_82583))
2817 		E1000_WRITE_REG(hw, E1000_RCTL, rctl & ~E1000_RCTL_EN);
2818 
2819 	/* Setup the Receive Control Register */
2820 	rctl &= ~(3 << E1000_RCTL_MO_SHIFT);
2821 	rctl |= E1000_RCTL_EN | E1000_RCTL_BAM |
2822 	    E1000_RCTL_LBM_NO | E1000_RCTL_RDMTS_HALF |
2823 	    (hw->mac.mc_filter_type << E1000_RCTL_MO_SHIFT);
2824 
2825 	/* Do not store bad packets */
2826 	rctl &= ~E1000_RCTL_SBP;
2827 
2828 	/* Enable Long Packet receive */
2829 	if (if_getmtu(ifp) > ETHERMTU)
2830 		rctl |= E1000_RCTL_LPE;
2831 	else
2832 		rctl &= ~E1000_RCTL_LPE;
2833 
2834         /* Strip the CRC */
2835         if (!em_disable_crc_stripping)
2836 		rctl |= E1000_RCTL_SECRC;
2837 
2838 	if (adapter->hw.mac.type >= e1000_82540) {
2839 		E1000_WRITE_REG(&adapter->hw, E1000_RADV,
2840 				adapter->rx_abs_int_delay.value);
2841 
2842 		/*
2843 		 * Set the interrupt throttling rate. Value is calculated
2844 		 * as DEFAULT_ITR = 1/(MAX_INTS_PER_SEC * 256ns)
2845 		 */
2846 		E1000_WRITE_REG(hw, E1000_ITR, DEFAULT_ITR);
2847 	}
2848 	E1000_WRITE_REG(&adapter->hw, E1000_RDTR,
2849 	    adapter->rx_int_delay.value);
2850 
2851 	/* Use extended rx descriptor formats */
2852 	rfctl = E1000_READ_REG(hw, E1000_RFCTL);
2853 	rfctl |= E1000_RFCTL_EXTEN;
2854 	/*
2855 	** When using MSIX interrupts we need to throttle
2856 	** using the EITR register (82574 only)
2857 	*/
2858 	if (hw->mac.type == e1000_82574) {
2859 		for (int i = 0; i < 4; i++)
2860 			E1000_WRITE_REG(hw, E1000_EITR_82574(i),
2861 			    DEFAULT_ITR);
2862 		/* Disable accelerated acknowledge */
2863 		rfctl |= E1000_RFCTL_ACK_DIS;
2864 	}
2865 	E1000_WRITE_REG(hw, E1000_RFCTL, rfctl);
2866 
2867 	rxcsum = E1000_READ_REG(hw, E1000_RXCSUM);
2868 	if (if_getcapenable(ifp) & IFCAP_RXCSUM &&
2869 	    adapter->hw.mac.type >= e1000_82543) {
2870 		if (adapter->tx_num_queues > 1) {
2871 			if (adapter->hw.mac.type >= igb_mac_min) {
2872 				rxcsum |= E1000_RXCSUM_PCSD;
2873 				if (hw->mac.type != e1000_82575)
2874 					rxcsum |= E1000_RXCSUM_CRCOFL;
2875 			} else
2876 				rxcsum |= E1000_RXCSUM_TUOFL |
2877 					E1000_RXCSUM_IPOFL |
2878 					E1000_RXCSUM_PCSD;
2879 		} else {
2880 			if (adapter->hw.mac.type >= igb_mac_min)
2881 				rxcsum |= E1000_RXCSUM_IPPCSE;
2882 			else
2883 				rxcsum |= E1000_RXCSUM_TUOFL | E1000_RXCSUM_IPOFL;
2884 			if (adapter->hw.mac.type > e1000_82575)
2885 				rxcsum |= E1000_RXCSUM_CRCOFL;
2886 		}
2887 	} else
2888 		rxcsum &= ~E1000_RXCSUM_TUOFL;
2889 
2890 	E1000_WRITE_REG(hw, E1000_RXCSUM, rxcsum);
2891 
2892 	if (adapter->rx_num_queues > 1) {
2893 		if (adapter->hw.mac.type >= igb_mac_min)
2894 			igb_initialize_rss_mapping(adapter);
2895 		else
2896 			em_initialize_rss_mapping(adapter);
2897 	}
2898 
2899 	/*
2900 	** XXX TEMPORARY WORKAROUND: on some systems with 82573
2901 	** long latencies are observed, like Lenovo X60. This
2902 	** change eliminates the problem, but since having positive
2903 	** values in RDTR is a known source of problems on other
2904 	** platforms another solution is being sought.
2905 	*/
2906 	if (hw->mac.type == e1000_82573)
2907 		E1000_WRITE_REG(hw, E1000_RDTR, 0x20);
2908 
2909 	for (i = 0, que = adapter->rx_queues; i < adapter->rx_num_queues; i++, que++) {
2910 	        struct rx_ring *rxr = &que->rxr;
2911 		/* Setup the Base and Length of the Rx Descriptor Ring */
2912 		u64 bus_addr = rxr->rx_paddr;
2913 #if 0
2914 		u32 rdt = adapter->rx_num_queues -1;  /* default */
2915 #endif
2916 
2917 		E1000_WRITE_REG(hw, E1000_RDLEN(i),
2918 		    scctx->isc_nrxd[0] * sizeof(union e1000_rx_desc_extended));
2919 		E1000_WRITE_REG(hw, E1000_RDBAH(i), (u32)(bus_addr >> 32));
2920 		E1000_WRITE_REG(hw, E1000_RDBAL(i), (u32)bus_addr);
2921 		/* Setup the Head and Tail Descriptor Pointers */
2922 		E1000_WRITE_REG(hw, E1000_RDH(i), 0);
2923 		E1000_WRITE_REG(hw, E1000_RDT(i), 0);
2924 	}
2925 
2926 	/*
2927 	 * Set PTHRESH for improved jumbo performance
2928 	 * According to 10.2.5.11 of Intel 82574 Datasheet,
2929 	 * RXDCTL(1) is written whenever RXDCTL(0) is written.
2930 	 * Only write to RXDCTL(1) if there is a need for different
2931 	 * settings.
2932 	 */
2933 
2934 	if (((adapter->hw.mac.type == e1000_ich9lan) ||
2935 	    (adapter->hw.mac.type == e1000_pch2lan) ||
2936 	    (adapter->hw.mac.type == e1000_ich10lan)) &&
2937 	    (if_getmtu(ifp) > ETHERMTU)) {
2938 		u32 rxdctl = E1000_READ_REG(hw, E1000_RXDCTL(0));
2939 		E1000_WRITE_REG(hw, E1000_RXDCTL(0), rxdctl | 3);
2940 	} else if (adapter->hw.mac.type == e1000_82574) {
2941 		for (int i = 0; i < adapter->rx_num_queues; i++) {
2942 			u32 rxdctl = E1000_READ_REG(hw, E1000_RXDCTL(i));
2943 			rxdctl |= 0x20; /* PTHRESH */
2944 			rxdctl |= 4 << 8; /* HTHRESH */
2945 			rxdctl |= 4 << 16;/* WTHRESH */
2946 			rxdctl |= 1 << 24; /* Switch to granularity */
2947 			E1000_WRITE_REG(hw, E1000_RXDCTL(i), rxdctl);
2948 		}
2949 	} else if (adapter->hw.mac.type >= igb_mac_min) {
2950 		u32 psize, srrctl = 0;
2951 
2952 		if (ifp->if_mtu > ETHERMTU) {
2953 			rctl |= E1000_RCTL_LPE;
2954 
2955 			/* Set maximum packet len */
2956 			psize = scctx->isc_max_frame_size;
2957 			if (psize <= 4096) {
2958 				srrctl |= 4096 >> E1000_SRRCTL_BSIZEPKT_SHIFT;
2959 				rctl |= E1000_RCTL_SZ_4096 | E1000_RCTL_BSEX;
2960 			} else if (psize > 4096) {
2961 				srrctl |= 8192 >> E1000_SRRCTL_BSIZEPKT_SHIFT;
2962 				rctl |= E1000_RCTL_SZ_8192 | E1000_RCTL_BSEX;
2963 			}
2964 
2965 			/* are we on a vlan? */
2966 			if (ifp->if_vlantrunk != NULL)
2967 				psize += VLAN_TAG_SIZE;
2968 			E1000_WRITE_REG(&adapter->hw, E1000_RLPML, psize);
2969 		} else {
2970 			rctl &= ~E1000_RCTL_LPE;
2971 			srrctl |= 2048 >> E1000_SRRCTL_BSIZEPKT_SHIFT;
2972 			rctl |= E1000_RCTL_SZ_2048;
2973 		}
2974 
2975 		/*
2976 		 * If TX flow control is disabled and there's >1 queue defined,
2977 		 * enable DROP.
2978 		 *
2979 		 * This drops frames rather than hanging the RX MAC for all queues.
2980 		 */
2981 		if ((adapter->rx_num_queues > 1) &&
2982 		    (adapter->fc == e1000_fc_none ||
2983 		     adapter->fc == e1000_fc_rx_pause)) {
2984 			srrctl |= E1000_SRRCTL_DROP_EN;
2985 		}
2986 			/* Setup the Base and Length of the Rx Descriptor Rings */
2987 		for (i = 0, que = adapter->rx_queues; i < adapter->rx_num_queues; i++, que++) {
2988 			struct rx_ring *rxr = &que->rxr;
2989 			u64 bus_addr = rxr->rx_paddr;
2990 			u32 rxdctl;
2991 
2992 #ifdef notyet
2993 			/* Configure for header split? -- ignore for now */
2994 			rxr->hdr_split = igb_header_split;
2995 #else
2996 			srrctl |= E1000_SRRCTL_DESCTYPE_ADV_ONEBUF;
2997 #endif
2998 
2999 
3000 			E1000_WRITE_REG(hw, E1000_RDLEN(i),
3001 					scctx->isc_nrxd[0] * sizeof(struct e1000_rx_desc));
3002 			E1000_WRITE_REG(hw, E1000_RDBAH(i),
3003 					(uint32_t)(bus_addr >> 32));
3004 			E1000_WRITE_REG(hw, E1000_RDBAL(i),
3005 					(uint32_t)bus_addr);
3006 			E1000_WRITE_REG(hw, E1000_SRRCTL(i), srrctl);
3007 			/* Enable this Queue */
3008 			rxdctl = E1000_READ_REG(hw, E1000_RXDCTL(i));
3009 			rxdctl |= E1000_RXDCTL_QUEUE_ENABLE;
3010 			rxdctl &= 0xFFF00000;
3011 			rxdctl |= IGB_RX_PTHRESH;
3012 			rxdctl |= IGB_RX_HTHRESH << 8;
3013 			rxdctl |= IGB_RX_WTHRESH << 16;
3014 			E1000_WRITE_REG(hw, E1000_RXDCTL(i), rxdctl);
3015 		}
3016 	}
3017 	if (adapter->hw.mac.type >= e1000_pch2lan) {
3018 		if (if_getmtu(ifp) > ETHERMTU)
3019 			e1000_lv_jumbo_workaround_ich8lan(hw, TRUE);
3020 		else
3021 			e1000_lv_jumbo_workaround_ich8lan(hw, FALSE);
3022 	}
3023 
3024         /* Make sure VLAN Filters are off */
3025         rctl &= ~E1000_RCTL_VFE;
3026 
3027 	if (adapter->rx_mbuf_sz == MCLBYTES)
3028 		rctl |= E1000_RCTL_SZ_2048;
3029 	else if (adapter->rx_mbuf_sz == MJUMPAGESIZE)
3030 		rctl |= E1000_RCTL_SZ_4096 | E1000_RCTL_BSEX;
3031 	else if (adapter->rx_mbuf_sz > MJUMPAGESIZE)
3032 		rctl |= E1000_RCTL_SZ_8192 | E1000_RCTL_BSEX;
3033 
3034 	/* ensure we clear use DTYPE of 00 here */
3035 	rctl &= ~0x00000C00;
3036 	/* Write out the settings */
3037 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
3038 
3039 	return;
3040 }
3041 
3042 static void
3043 em_if_vlan_register(if_ctx_t ctx, u16 vtag)
3044 {
3045         struct adapter  *adapter = iflib_get_softc(ctx);
3046 	u32		index, bit;
3047 
3048 	index = (vtag >> 5) & 0x7F;
3049 	bit = vtag & 0x1F;
3050 	adapter->shadow_vfta[index] |= (1 << bit);
3051 	++adapter->num_vlans;
3052 }
3053 
3054 static void
3055 em_if_vlan_unregister(if_ctx_t ctx, u16 vtag)
3056 {
3057         struct adapter	*adapter = iflib_get_softc(ctx);
3058 	u32		index, bit;
3059 
3060 	index = (vtag >> 5) & 0x7F;
3061 	bit = vtag & 0x1F;
3062 	adapter->shadow_vfta[index] &= ~(1 << bit);
3063 	--adapter->num_vlans;
3064 }
3065 
3066 static void
3067 em_setup_vlan_hw_support(struct adapter *adapter)
3068 {
3069 	struct e1000_hw *hw = &adapter->hw;
3070 	u32             reg;
3071 
3072 	/*
3073 	** We get here thru init_locked, meaning
3074 	** a soft reset, this has already cleared
3075 	** the VFTA and other state, so if there
3076 	** have been no vlan's registered do nothing.
3077 	*/
3078 	if (adapter->num_vlans == 0)
3079                 return;
3080 
3081 	/*
3082 	** A soft reset zero's out the VFTA, so
3083 	** we need to repopulate it now.
3084 	*/
3085 	for (int i = 0; i < EM_VFTA_SIZE; i++)
3086                 if (adapter->shadow_vfta[i] != 0)
3087 			E1000_WRITE_REG_ARRAY(hw, E1000_VFTA,
3088                             i, adapter->shadow_vfta[i]);
3089 
3090 	reg = E1000_READ_REG(hw, E1000_CTRL);
3091 	reg |= E1000_CTRL_VME;
3092 	E1000_WRITE_REG(hw, E1000_CTRL, reg);
3093 
3094 	/* Enable the Filter Table */
3095 	reg = E1000_READ_REG(hw, E1000_RCTL);
3096 	reg &= ~E1000_RCTL_CFIEN;
3097 	reg |= E1000_RCTL_VFE;
3098 	E1000_WRITE_REG(hw, E1000_RCTL, reg);
3099 }
3100 
3101 static void
3102 em_if_enable_intr(if_ctx_t ctx)
3103 {
3104         struct adapter *adapter = iflib_get_softc(ctx);
3105 	struct e1000_hw *hw = &adapter->hw;
3106 	u32 ims_mask = IMS_ENABLE_MASK;
3107 
3108 	if (hw->mac.type == e1000_82574) {
3109 		E1000_WRITE_REG(hw, EM_EIAC, adapter->ims);
3110 		ims_mask |= adapter->ims;
3111 	} if (adapter->intr_type == IFLIB_INTR_MSIX && hw->mac.type >= igb_mac_min)  {
3112 		u32 mask = (adapter->que_mask | adapter->link_mask);
3113 
3114 		E1000_WRITE_REG(&adapter->hw, E1000_EIAC, mask);
3115 		E1000_WRITE_REG(&adapter->hw, E1000_EIAM, mask);
3116 		E1000_WRITE_REG(&adapter->hw, E1000_EIMS, mask);
3117 		ims_mask = E1000_IMS_LSC;
3118 	}
3119 
3120 	E1000_WRITE_REG(hw, E1000_IMS, ims_mask);
3121 }
3122 
3123 static void
3124 em_if_disable_intr(if_ctx_t ctx)
3125 {
3126         struct adapter *adapter = iflib_get_softc(ctx);
3127 	struct e1000_hw *hw = &adapter->hw;
3128 
3129 	if (adapter->intr_type == IFLIB_INTR_MSIX) {
3130 		if (hw->mac.type >= igb_mac_min)
3131 			E1000_WRITE_REG(&adapter->hw, E1000_EIMC, ~0);
3132 		E1000_WRITE_REG(&adapter->hw, E1000_EIAC, 0);
3133 	}
3134 	E1000_WRITE_REG(&adapter->hw, E1000_IMC, 0xffffffff);
3135 }
3136 
3137 /*
3138  * Bit of a misnomer, what this really means is
3139  * to enable OS management of the system... aka
3140  * to disable special hardware management features
3141  */
3142 static void
3143 em_init_manageability(struct adapter *adapter)
3144 {
3145 	/* A shared code workaround */
3146 #define E1000_82542_MANC2H E1000_MANC2H
3147 	if (adapter->has_manage) {
3148 		int manc2h = E1000_READ_REG(&adapter->hw, E1000_MANC2H);
3149 		int manc = E1000_READ_REG(&adapter->hw, E1000_MANC);
3150 
3151 		/* disable hardware interception of ARP */
3152 		manc &= ~(E1000_MANC_ARP_EN);
3153 
3154                 /* enable receiving management packets to the host */
3155 		manc |= E1000_MANC_EN_MNG2HOST;
3156 #define E1000_MNG2HOST_PORT_623 (1 << 5)
3157 #define E1000_MNG2HOST_PORT_664 (1 << 6)
3158 		manc2h |= E1000_MNG2HOST_PORT_623;
3159 		manc2h |= E1000_MNG2HOST_PORT_664;
3160 		E1000_WRITE_REG(&adapter->hw, E1000_MANC2H, manc2h);
3161 		E1000_WRITE_REG(&adapter->hw, E1000_MANC, manc);
3162 	}
3163 }
3164 
3165 /*
3166  * Give control back to hardware management
3167  * controller if there is one.
3168  */
3169 static void
3170 em_release_manageability(struct adapter *adapter)
3171 {
3172 	if (adapter->has_manage) {
3173 		int manc = E1000_READ_REG(&adapter->hw, E1000_MANC);
3174 
3175 		/* re-enable hardware interception of ARP */
3176 		manc |= E1000_MANC_ARP_EN;
3177 		manc &= ~E1000_MANC_EN_MNG2HOST;
3178 
3179 		E1000_WRITE_REG(&adapter->hw, E1000_MANC, manc);
3180 	}
3181 }
3182 
3183 /*
3184  * em_get_hw_control sets the {CTRL_EXT|FWSM}:DRV_LOAD bit.
3185  * For ASF and Pass Through versions of f/w this means
3186  * that the driver is loaded. For AMT version type f/w
3187  * this means that the network i/f is open.
3188  */
3189 static void
3190 em_get_hw_control(struct adapter *adapter)
3191 {
3192 	u32 ctrl_ext, swsm;
3193 
3194 	if (adapter->hw.mac.type == e1000_82573) {
3195 		swsm = E1000_READ_REG(&adapter->hw, E1000_SWSM);
3196 		E1000_WRITE_REG(&adapter->hw, E1000_SWSM,
3197 		    swsm | E1000_SWSM_DRV_LOAD);
3198 		return;
3199 	}
3200 	/* else */
3201 	ctrl_ext = E1000_READ_REG(&adapter->hw, E1000_CTRL_EXT);
3202 	E1000_WRITE_REG(&adapter->hw, E1000_CTRL_EXT,
3203 	    ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
3204 	return;
3205 }
3206 
3207 /*
3208  * em_release_hw_control resets {CTRL_EXT|FWSM}:DRV_LOAD bit.
3209  * For ASF and Pass Through versions of f/w this means that
3210  * the driver is no longer loaded. For AMT versions of the
3211  * f/w this means that the network i/f is closed.
3212  */
3213 static void
3214 em_release_hw_control(struct adapter *adapter)
3215 {
3216 	u32 ctrl_ext, swsm;
3217 
3218 	if (!adapter->has_manage)
3219 		return;
3220 
3221 	if (adapter->hw.mac.type == e1000_82573) {
3222 		swsm = E1000_READ_REG(&adapter->hw, E1000_SWSM);
3223 		E1000_WRITE_REG(&adapter->hw, E1000_SWSM,
3224 		    swsm & ~E1000_SWSM_DRV_LOAD);
3225 		return;
3226 	}
3227 	/* else */
3228 	ctrl_ext = E1000_READ_REG(&adapter->hw, E1000_CTRL_EXT);
3229 	E1000_WRITE_REG(&adapter->hw, E1000_CTRL_EXT,
3230 	    ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
3231 	return;
3232 }
3233 
3234 static int
3235 em_is_valid_ether_addr(u8 *addr)
3236 {
3237 	char zero_addr[6] = { 0, 0, 0, 0, 0, 0 };
3238 
3239 	if ((addr[0] & 1) || (!bcmp(addr, zero_addr, ETHER_ADDR_LEN))) {
3240 		return (FALSE);
3241 	}
3242 
3243 	return (TRUE);
3244 }
3245 
3246 /*
3247 ** Parse the interface capabilities with regard
3248 ** to both system management and wake-on-lan for
3249 ** later use.
3250 */
3251 static void
3252 em_get_wakeup(if_ctx_t ctx)
3253 {
3254 	struct adapter	*adapter = iflib_get_softc(ctx);
3255 	device_t dev = iflib_get_dev(ctx);
3256 	u16		eeprom_data = 0, device_id, apme_mask;
3257 
3258 	adapter->has_manage = e1000_enable_mng_pass_thru(&adapter->hw);
3259 	apme_mask = EM_EEPROM_APME;
3260 
3261 	switch (adapter->hw.mac.type) {
3262 	case e1000_82542:
3263 	case e1000_82543:
3264 		break;
3265 	case e1000_82544:
3266 		e1000_read_nvm(&adapter->hw,
3267 		    NVM_INIT_CONTROL2_REG, 1, &eeprom_data);
3268 		apme_mask = EM_82544_APME;
3269 		break;
3270 	case e1000_82546:
3271 	case e1000_82546_rev_3:
3272 		if (adapter->hw.bus.func == 1) {
3273 			e1000_read_nvm(&adapter->hw,
3274 			    NVM_INIT_CONTROL3_PORT_B, 1, &eeprom_data);
3275 			break;
3276 		} else
3277 			e1000_read_nvm(&adapter->hw,
3278 			    NVM_INIT_CONTROL3_PORT_A, 1, &eeprom_data);
3279 		break;
3280 	case e1000_82573:
3281 	case e1000_82583:
3282 		adapter->has_amt = TRUE;
3283 		/* Falls thru */
3284 	case e1000_82571:
3285 	case e1000_82572:
3286 	case e1000_80003es2lan:
3287 		if (adapter->hw.bus.func == 1) {
3288 			e1000_read_nvm(&adapter->hw,
3289 			    NVM_INIT_CONTROL3_PORT_B, 1, &eeprom_data);
3290 			break;
3291 		} else
3292 			e1000_read_nvm(&adapter->hw,
3293 			    NVM_INIT_CONTROL3_PORT_A, 1, &eeprom_data);
3294 		break;
3295 	case e1000_ich8lan:
3296 	case e1000_ich9lan:
3297 	case e1000_ich10lan:
3298 	case e1000_pchlan:
3299 	case e1000_pch2lan:
3300 		apme_mask = E1000_WUC_APME;
3301 		adapter->has_amt = TRUE;
3302 		eeprom_data = E1000_READ_REG(&adapter->hw, E1000_WUC);
3303 		break;
3304 	default:
3305 		e1000_read_nvm(&adapter->hw,
3306 		    NVM_INIT_CONTROL3_PORT_A, 1, &eeprom_data);
3307 		break;
3308 	}
3309 	if (eeprom_data & apme_mask)
3310 		adapter->wol = (E1000_WUFC_MAG | E1000_WUFC_MC);
3311 	/*
3312          * We have the eeprom settings, now apply the special cases
3313          * where the eeprom may be wrong or the board won't support
3314          * wake on lan on a particular port
3315 	 */
3316 	device_id = pci_get_device(dev);
3317         switch (device_id) {
3318 	case E1000_DEV_ID_82546GB_PCIE:
3319 		adapter->wol = 0;
3320 		break;
3321 	case E1000_DEV_ID_82546EB_FIBER:
3322 	case E1000_DEV_ID_82546GB_FIBER:
3323 		/* Wake events only supported on port A for dual fiber
3324 		 * regardless of eeprom setting */
3325 		if (E1000_READ_REG(&adapter->hw, E1000_STATUS) &
3326 		    E1000_STATUS_FUNC_1)
3327 			adapter->wol = 0;
3328 		break;
3329 	case E1000_DEV_ID_82546GB_QUAD_COPPER_KSP3:
3330                 /* if quad port adapter, disable WoL on all but port A */
3331 		if (global_quad_port_a != 0)
3332 			adapter->wol = 0;
3333 		/* Reset for multiple quad port adapters */
3334 		if (++global_quad_port_a == 4)
3335 			global_quad_port_a = 0;
3336                 break;
3337 	case E1000_DEV_ID_82571EB_FIBER:
3338 		/* Wake events only supported on port A for dual fiber
3339 		 * regardless of eeprom setting */
3340 		if (E1000_READ_REG(&adapter->hw, E1000_STATUS) &
3341 		    E1000_STATUS_FUNC_1)
3342 			adapter->wol = 0;
3343 		break;
3344 	case E1000_DEV_ID_82571EB_QUAD_COPPER:
3345 	case E1000_DEV_ID_82571EB_QUAD_FIBER:
3346 	case E1000_DEV_ID_82571EB_QUAD_COPPER_LP:
3347                 /* if quad port adapter, disable WoL on all but port A */
3348 		if (global_quad_port_a != 0)
3349 			adapter->wol = 0;
3350 		/* Reset for multiple quad port adapters */
3351 		if (++global_quad_port_a == 4)
3352 			global_quad_port_a = 0;
3353                 break;
3354 	}
3355 	return;
3356 }
3357 
3358 
3359 /*
3360  * Enable PCI Wake On Lan capability
3361  */
3362 static void
3363 em_enable_wakeup(if_ctx_t ctx)
3364 {
3365 	struct adapter	*adapter = iflib_get_softc(ctx);
3366 	device_t dev = iflib_get_dev(ctx);
3367 	if_t ifp = iflib_get_ifp(ctx);
3368 	u32		pmc, ctrl, ctrl_ext, rctl;
3369 	u16     	status;
3370 
3371 	if ((pci_find_cap(dev, PCIY_PMG, &pmc) != 0))
3372 		return;
3373 
3374 	/* Advertise the wakeup capability */
3375 	ctrl = E1000_READ_REG(&adapter->hw, E1000_CTRL);
3376 	ctrl |= (E1000_CTRL_SWDPIN2 | E1000_CTRL_SWDPIN3);
3377 	E1000_WRITE_REG(&adapter->hw, E1000_CTRL, ctrl);
3378 	E1000_WRITE_REG(&adapter->hw, E1000_WUC, E1000_WUC_PME_EN);
3379 
3380 	if ((adapter->hw.mac.type == e1000_ich8lan) ||
3381 	    (adapter->hw.mac.type == e1000_pchlan) ||
3382 	    (adapter->hw.mac.type == e1000_ich9lan) ||
3383 	    (adapter->hw.mac.type == e1000_ich10lan))
3384 		e1000_suspend_workarounds_ich8lan(&adapter->hw);
3385 
3386 	/* Keep the laser running on Fiber adapters */
3387 	if (adapter->hw.phy.media_type == e1000_media_type_fiber ||
3388 	    adapter->hw.phy.media_type == e1000_media_type_internal_serdes) {
3389 		ctrl_ext = E1000_READ_REG(&adapter->hw, E1000_CTRL_EXT);
3390 		ctrl_ext |= E1000_CTRL_EXT_SDP3_DATA;
3391 		E1000_WRITE_REG(&adapter->hw, E1000_CTRL_EXT, ctrl_ext);
3392 	}
3393 
3394 	/*
3395 	** Determine type of Wakeup: note that wol
3396 	** is set with all bits on by default.
3397 	*/
3398 	if ((if_getcapenable(ifp) & IFCAP_WOL_MAGIC) == 0)
3399 		adapter->wol &= ~E1000_WUFC_MAG;
3400 
3401 	if ((if_getcapenable(ifp) & IFCAP_WOL_MCAST) == 0)
3402 		adapter->wol &= ~E1000_WUFC_MC;
3403 	else {
3404 		rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
3405 		rctl |= E1000_RCTL_MPE;
3406 		E1000_WRITE_REG(&adapter->hw, E1000_RCTL, rctl);
3407 	}
3408 
3409 	if ((adapter->hw.mac.type == e1000_pchlan) ||
3410 	    (adapter->hw.mac.type == e1000_pch2lan)) {
3411 		if (em_enable_phy_wakeup(adapter))
3412 			return;
3413 	} else {
3414 		E1000_WRITE_REG(&adapter->hw, E1000_WUC, E1000_WUC_PME_EN);
3415 		E1000_WRITE_REG(&adapter->hw, E1000_WUFC, adapter->wol);
3416 	}
3417 
3418 	if (adapter->hw.phy.type == e1000_phy_igp_3)
3419 		e1000_igp3_phy_powerdown_workaround_ich8lan(&adapter->hw);
3420 
3421         /* Request PME */
3422         status = pci_read_config(dev, pmc + PCIR_POWER_STATUS, 2);
3423 	status &= ~(PCIM_PSTAT_PME | PCIM_PSTAT_PMEENABLE);
3424 	if (if_getcapenable(ifp) & IFCAP_WOL)
3425 		status |= PCIM_PSTAT_PME | PCIM_PSTAT_PMEENABLE;
3426         pci_write_config(dev, pmc + PCIR_POWER_STATUS, status, 2);
3427 
3428 	return;
3429 }
3430 
3431 /*
3432 ** WOL in the newer chipset interfaces (pchlan)
3433 ** require thing to be copied into the phy
3434 */
3435 static int
3436 em_enable_phy_wakeup(struct adapter *adapter)
3437 {
3438 	struct e1000_hw *hw = &adapter->hw;
3439 	u32 mreg, ret = 0;
3440 	u16 preg;
3441 
3442 	/* copy MAC RARs to PHY RARs */
3443 	e1000_copy_rx_addrs_to_phy_ich8lan(hw);
3444 
3445 	/* copy MAC MTA to PHY MTA */
3446 	for (int i = 0; i < adapter->hw.mac.mta_reg_count; i++) {
3447 		mreg = E1000_READ_REG_ARRAY(hw, E1000_MTA, i);
3448 		e1000_write_phy_reg(hw, BM_MTA(i), (u16)(mreg & 0xFFFF));
3449 		e1000_write_phy_reg(hw, BM_MTA(i) + 1,
3450 		    (u16)((mreg >> 16) & 0xFFFF));
3451 	}
3452 
3453 	/* configure PHY Rx Control register */
3454 	e1000_read_phy_reg(&adapter->hw, BM_RCTL, &preg);
3455 	mreg = E1000_READ_REG(hw, E1000_RCTL);
3456 	if (mreg & E1000_RCTL_UPE)
3457 		preg |= BM_RCTL_UPE;
3458 	if (mreg & E1000_RCTL_MPE)
3459 		preg |= BM_RCTL_MPE;
3460 	preg &= ~(BM_RCTL_MO_MASK);
3461 	if (mreg & E1000_RCTL_MO_3)
3462 		preg |= (((mreg & E1000_RCTL_MO_3) >> E1000_RCTL_MO_SHIFT)
3463 				<< BM_RCTL_MO_SHIFT);
3464 	if (mreg & E1000_RCTL_BAM)
3465 		preg |= BM_RCTL_BAM;
3466 	if (mreg & E1000_RCTL_PMCF)
3467 		preg |= BM_RCTL_PMCF;
3468 	mreg = E1000_READ_REG(hw, E1000_CTRL);
3469 	if (mreg & E1000_CTRL_RFCE)
3470 		preg |= BM_RCTL_RFCE;
3471 	e1000_write_phy_reg(&adapter->hw, BM_RCTL, preg);
3472 
3473 	/* enable PHY wakeup in MAC register */
3474 	E1000_WRITE_REG(hw, E1000_WUC,
3475 	    E1000_WUC_PHY_WAKE | E1000_WUC_PME_EN);
3476 	E1000_WRITE_REG(hw, E1000_WUFC, adapter->wol);
3477 
3478 	/* configure and enable PHY wakeup in PHY registers */
3479 	e1000_write_phy_reg(&adapter->hw, BM_WUFC, adapter->wol);
3480 	e1000_write_phy_reg(&adapter->hw, BM_WUC, E1000_WUC_PME_EN);
3481 
3482 	/* activate PHY wakeup */
3483 	ret = hw->phy.ops.acquire(hw);
3484 	if (ret) {
3485 		printf("Could not acquire PHY\n");
3486 		return ret;
3487 	}
3488 	e1000_write_phy_reg_mdic(hw, IGP01E1000_PHY_PAGE_SELECT,
3489 	                         (BM_WUC_ENABLE_PAGE << IGP_PAGE_SHIFT));
3490 	ret = e1000_read_phy_reg_mdic(hw, BM_WUC_ENABLE_REG, &preg);
3491 	if (ret) {
3492 		printf("Could not read PHY page 769\n");
3493 		goto out;
3494 	}
3495 	preg |= BM_WUC_ENABLE_BIT | BM_WUC_HOST_WU_BIT;
3496 	ret = e1000_write_phy_reg_mdic(hw, BM_WUC_ENABLE_REG, preg);
3497 	if (ret)
3498 		printf("Could not set PHY Host Wakeup bit\n");
3499 out:
3500 	hw->phy.ops.release(hw);
3501 
3502 	return ret;
3503 }
3504 
3505 static void
3506 em_if_led_func(if_ctx_t ctx, int onoff)
3507 {
3508 	struct adapter	*adapter = iflib_get_softc(ctx);
3509 
3510 	if (onoff) {
3511 		e1000_setup_led(&adapter->hw);
3512 		e1000_led_on(&adapter->hw);
3513 	} else {
3514 		e1000_led_off(&adapter->hw);
3515 		e1000_cleanup_led(&adapter->hw);
3516 	}
3517 }
3518 
3519 /*
3520 ** Disable the L0S and L1 LINK states
3521 */
3522 static void
3523 em_disable_aspm(struct adapter *adapter)
3524 {
3525 	int		base, reg;
3526 	u16		link_cap,link_ctrl;
3527 	device_t	dev = adapter->dev;
3528 
3529 	switch (adapter->hw.mac.type) {
3530 		case e1000_82573:
3531 		case e1000_82574:
3532 		case e1000_82583:
3533 			break;
3534 		default:
3535 			return;
3536 	}
3537 	if (pci_find_cap(dev, PCIY_EXPRESS, &base) != 0)
3538 		return;
3539 	reg = base + PCIER_LINK_CAP;
3540 	link_cap = pci_read_config(dev, reg, 2);
3541 	if ((link_cap & PCIEM_LINK_CAP_ASPM) == 0)
3542 		return;
3543 	reg = base + PCIER_LINK_CTL;
3544 	link_ctrl = pci_read_config(dev, reg, 2);
3545 	link_ctrl &= ~PCIEM_LINK_CTL_ASPMC;
3546 	pci_write_config(dev, reg, link_ctrl, 2);
3547 	return;
3548 }
3549 
3550 /**********************************************************************
3551  *
3552  *  Update the board statistics counters.
3553  *
3554  **********************************************************************/
3555 static void
3556 em_update_stats_counters(struct adapter *adapter)
3557 {
3558 
3559 	if(adapter->hw.phy.media_type == e1000_media_type_copper ||
3560 	   (E1000_READ_REG(&adapter->hw, E1000_STATUS) & E1000_STATUS_LU)) {
3561 		adapter->stats.symerrs += E1000_READ_REG(&adapter->hw, E1000_SYMERRS);
3562 		adapter->stats.sec += E1000_READ_REG(&adapter->hw, E1000_SEC);
3563 	}
3564 	adapter->stats.crcerrs += E1000_READ_REG(&adapter->hw, E1000_CRCERRS);
3565 	adapter->stats.mpc += E1000_READ_REG(&adapter->hw, E1000_MPC);
3566 	adapter->stats.scc += E1000_READ_REG(&adapter->hw, E1000_SCC);
3567 	adapter->stats.ecol += E1000_READ_REG(&adapter->hw, E1000_ECOL);
3568 
3569 	adapter->stats.mcc += E1000_READ_REG(&adapter->hw, E1000_MCC);
3570 	adapter->stats.latecol += E1000_READ_REG(&adapter->hw, E1000_LATECOL);
3571 	adapter->stats.colc += E1000_READ_REG(&adapter->hw, E1000_COLC);
3572 	adapter->stats.dc += E1000_READ_REG(&adapter->hw, E1000_DC);
3573 	adapter->stats.rlec += E1000_READ_REG(&adapter->hw, E1000_RLEC);
3574 	adapter->stats.xonrxc += E1000_READ_REG(&adapter->hw, E1000_XONRXC);
3575 	adapter->stats.xontxc += E1000_READ_REG(&adapter->hw, E1000_XONTXC);
3576 	adapter->stats.xoffrxc += E1000_READ_REG(&adapter->hw, E1000_XOFFRXC);
3577 	adapter->stats.xofftxc += E1000_READ_REG(&adapter->hw, E1000_XOFFTXC);
3578 	adapter->stats.fcruc += E1000_READ_REG(&adapter->hw, E1000_FCRUC);
3579 	adapter->stats.prc64 += E1000_READ_REG(&adapter->hw, E1000_PRC64);
3580 	adapter->stats.prc127 += E1000_READ_REG(&adapter->hw, E1000_PRC127);
3581 	adapter->stats.prc255 += E1000_READ_REG(&adapter->hw, E1000_PRC255);
3582 	adapter->stats.prc511 += E1000_READ_REG(&adapter->hw, E1000_PRC511);
3583 	adapter->stats.prc1023 += E1000_READ_REG(&adapter->hw, E1000_PRC1023);
3584 	adapter->stats.prc1522 += E1000_READ_REG(&adapter->hw, E1000_PRC1522);
3585 	adapter->stats.gprc += E1000_READ_REG(&adapter->hw, E1000_GPRC);
3586 	adapter->stats.bprc += E1000_READ_REG(&adapter->hw, E1000_BPRC);
3587 	adapter->stats.mprc += E1000_READ_REG(&adapter->hw, E1000_MPRC);
3588 	adapter->stats.gptc += E1000_READ_REG(&adapter->hw, E1000_GPTC);
3589 
3590 	/* For the 64-bit byte counters the low dword must be read first. */
3591 	/* Both registers clear on the read of the high dword */
3592 
3593 	adapter->stats.gorc += E1000_READ_REG(&adapter->hw, E1000_GORCL) +
3594 	    ((u64)E1000_READ_REG(&adapter->hw, E1000_GORCH) << 32);
3595 	adapter->stats.gotc += E1000_READ_REG(&adapter->hw, E1000_GOTCL) +
3596 	    ((u64)E1000_READ_REG(&adapter->hw, E1000_GOTCH) << 32);
3597 
3598 	adapter->stats.rnbc += E1000_READ_REG(&adapter->hw, E1000_RNBC);
3599 	adapter->stats.ruc += E1000_READ_REG(&adapter->hw, E1000_RUC);
3600 	adapter->stats.rfc += E1000_READ_REG(&adapter->hw, E1000_RFC);
3601 	adapter->stats.roc += E1000_READ_REG(&adapter->hw, E1000_ROC);
3602 	adapter->stats.rjc += E1000_READ_REG(&adapter->hw, E1000_RJC);
3603 
3604 	adapter->stats.tor += E1000_READ_REG(&adapter->hw, E1000_TORH);
3605 	adapter->stats.tot += E1000_READ_REG(&adapter->hw, E1000_TOTH);
3606 
3607 	adapter->stats.tpr += E1000_READ_REG(&adapter->hw, E1000_TPR);
3608 	adapter->stats.tpt += E1000_READ_REG(&adapter->hw, E1000_TPT);
3609 	adapter->stats.ptc64 += E1000_READ_REG(&adapter->hw, E1000_PTC64);
3610 	adapter->stats.ptc127 += E1000_READ_REG(&adapter->hw, E1000_PTC127);
3611 	adapter->stats.ptc255 += E1000_READ_REG(&adapter->hw, E1000_PTC255);
3612 	adapter->stats.ptc511 += E1000_READ_REG(&adapter->hw, E1000_PTC511);
3613 	adapter->stats.ptc1023 += E1000_READ_REG(&adapter->hw, E1000_PTC1023);
3614 	adapter->stats.ptc1522 += E1000_READ_REG(&adapter->hw, E1000_PTC1522);
3615 	adapter->stats.mptc += E1000_READ_REG(&adapter->hw, E1000_MPTC);
3616 	adapter->stats.bptc += E1000_READ_REG(&adapter->hw, E1000_BPTC);
3617 
3618 	/* Interrupt Counts */
3619 
3620 	adapter->stats.iac += E1000_READ_REG(&adapter->hw, E1000_IAC);
3621 	adapter->stats.icrxptc += E1000_READ_REG(&adapter->hw, E1000_ICRXPTC);
3622 	adapter->stats.icrxatc += E1000_READ_REG(&adapter->hw, E1000_ICRXATC);
3623 	adapter->stats.ictxptc += E1000_READ_REG(&adapter->hw, E1000_ICTXPTC);
3624 	adapter->stats.ictxatc += E1000_READ_REG(&adapter->hw, E1000_ICTXATC);
3625 	adapter->stats.ictxqec += E1000_READ_REG(&adapter->hw, E1000_ICTXQEC);
3626 	adapter->stats.ictxqmtc += E1000_READ_REG(&adapter->hw, E1000_ICTXQMTC);
3627 	adapter->stats.icrxdmtc += E1000_READ_REG(&adapter->hw, E1000_ICRXDMTC);
3628 	adapter->stats.icrxoc += E1000_READ_REG(&adapter->hw, E1000_ICRXOC);
3629 
3630 	if (adapter->hw.mac.type >= e1000_82543) {
3631 		adapter->stats.algnerrc +=
3632 		E1000_READ_REG(&adapter->hw, E1000_ALGNERRC);
3633 		adapter->stats.rxerrc +=
3634 		E1000_READ_REG(&adapter->hw, E1000_RXERRC);
3635 		adapter->stats.tncrs +=
3636 		E1000_READ_REG(&adapter->hw, E1000_TNCRS);
3637 		adapter->stats.cexterr +=
3638 		E1000_READ_REG(&adapter->hw, E1000_CEXTERR);
3639 		adapter->stats.tsctc +=
3640 		E1000_READ_REG(&adapter->hw, E1000_TSCTC);
3641 		adapter->stats.tsctfc +=
3642 		E1000_READ_REG(&adapter->hw, E1000_TSCTFC);
3643 	}
3644 }
3645 
3646 static uint64_t
3647 em_if_get_counter(if_ctx_t ctx, ift_counter cnt)
3648 {
3649 	struct adapter *adapter = iflib_get_softc(ctx);
3650 	struct ifnet *ifp = iflib_get_ifp(ctx);
3651 
3652 	switch (cnt) {
3653 	case IFCOUNTER_COLLISIONS:
3654 		return (adapter->stats.colc);
3655 	case IFCOUNTER_IERRORS:
3656 		return (adapter->dropped_pkts + adapter->stats.rxerrc +
3657 		    adapter->stats.crcerrs + adapter->stats.algnerrc +
3658 		    adapter->stats.ruc + adapter->stats.roc +
3659 		    adapter->stats.mpc + adapter->stats.cexterr);
3660 	case IFCOUNTER_OERRORS:
3661 		return (adapter->stats.ecol + adapter->stats.latecol +
3662 		    adapter->watchdog_events);
3663 	default:
3664 		return (if_get_counter_default(ifp, cnt));
3665 	}
3666 }
3667 
3668 /* Export a single 32-bit register via a read-only sysctl. */
3669 static int
3670 em_sysctl_reg_handler(SYSCTL_HANDLER_ARGS)
3671 {
3672 	struct adapter *adapter;
3673 	u_int val;
3674 
3675 	adapter = oidp->oid_arg1;
3676 	val = E1000_READ_REG(&adapter->hw, oidp->oid_arg2);
3677 	return (sysctl_handle_int(oidp, &val, 0, req));
3678 }
3679 
3680 /*
3681  * Add sysctl variables, one per statistic, to the system.
3682  */
3683 static void
3684 em_add_hw_stats(struct adapter *adapter)
3685 {
3686 	device_t dev = iflib_get_dev(adapter->ctx);
3687         struct em_tx_queue *tx_que = adapter->tx_queues;
3688 	struct em_rx_queue *rx_que = adapter->rx_queues;
3689 
3690 	struct sysctl_ctx_list *ctx = device_get_sysctl_ctx(dev);
3691 	struct sysctl_oid *tree = device_get_sysctl_tree(dev);
3692 	struct sysctl_oid_list *child = SYSCTL_CHILDREN(tree);
3693 	struct e1000_hw_stats *stats = &adapter->stats;
3694 
3695 	struct sysctl_oid *stat_node, *queue_node, *int_node;
3696 	struct sysctl_oid_list *stat_list, *queue_list, *int_list;
3697 
3698 #define QUEUE_NAME_LEN 32
3699 	char namebuf[QUEUE_NAME_LEN];
3700 
3701 	/* Driver Statistics */
3702 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "dropped",
3703 			CTLFLAG_RD, &adapter->dropped_pkts,
3704 			"Driver dropped packets");
3705 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "link_irq",
3706 			CTLFLAG_RD, &adapter->link_irq,
3707 			"Link MSIX IRQ Handled");
3708 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "mbuf_defrag_fail",
3709 			 CTLFLAG_RD, &adapter->mbuf_defrag_failed,
3710 			 "Defragmenting mbuf chain failed");
3711 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "tx_dma_fail",
3712 			CTLFLAG_RD, &adapter->no_tx_dma_setup,
3713 			"Driver tx dma failure in xmit");
3714 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "rx_overruns",
3715 			CTLFLAG_RD, &adapter->rx_overruns,
3716 			"RX overruns");
3717 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "watchdog_timeouts",
3718 			CTLFLAG_RD, &adapter->watchdog_events,
3719 			"Watchdog timeouts");
3720 
3721 	SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "device_control",
3722 			CTLTYPE_UINT | CTLFLAG_RD, adapter, E1000_CTRL,
3723 			em_sysctl_reg_handler, "IU",
3724 			"Device Control Register");
3725 	SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "rx_control",
3726 			CTLTYPE_UINT | CTLFLAG_RD, adapter, E1000_RCTL,
3727 			em_sysctl_reg_handler, "IU",
3728 			"Receiver Control Register");
3729 	SYSCTL_ADD_UINT(ctx, child, OID_AUTO, "fc_high_water",
3730 			CTLFLAG_RD, &adapter->hw.fc.high_water, 0,
3731 			"Flow Control High Watermark");
3732 	SYSCTL_ADD_UINT(ctx, child, OID_AUTO, "fc_low_water",
3733 			CTLFLAG_RD, &adapter->hw.fc.low_water, 0,
3734 			"Flow Control Low Watermark");
3735 
3736 	for (int i = 0; i < adapter->tx_num_queues; i++, tx_que++) {
3737 	        struct tx_ring *txr = &tx_que->txr;
3738 		snprintf(namebuf, QUEUE_NAME_LEN, "queue_tx_%d", i);
3739 		queue_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, namebuf,
3740 					    CTLFLAG_RD, NULL, "TX Queue Name");
3741 		queue_list = SYSCTL_CHILDREN(queue_node);
3742 
3743 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "txd_head",
3744 				CTLTYPE_UINT | CTLFLAG_RD, adapter,
3745 				E1000_TDH(txr->me),
3746 				em_sysctl_reg_handler, "IU",
3747  				"Transmit Descriptor Head");
3748 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "txd_tail",
3749 				CTLTYPE_UINT | CTLFLAG_RD, adapter,
3750 				E1000_TDT(txr->me),
3751 				em_sysctl_reg_handler, "IU",
3752  				"Transmit Descriptor Tail");
3753 		SYSCTL_ADD_ULONG(ctx, queue_list, OID_AUTO, "tx_irq",
3754 				CTLFLAG_RD, &txr->tx_irq,
3755 				"Queue MSI-X Transmit Interrupts");
3756 		SYSCTL_ADD_ULONG(ctx, queue_list, OID_AUTO, "no_desc_avail",
3757 				CTLFLAG_RD, &txr->no_desc_avail,
3758 				"Queue No Descriptor Available");
3759 	}
3760 
3761 	for (int j = 0; j < adapter->rx_num_queues; j++, rx_que++) {
3762 	        struct rx_ring *rxr = &rx_que->rxr;
3763 		snprintf(namebuf, QUEUE_NAME_LEN, "queue_rx_%d", j);
3764 		queue_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, namebuf,
3765 					    CTLFLAG_RD, NULL, "RX Queue Name");
3766 		queue_list = SYSCTL_CHILDREN(queue_node);
3767 
3768 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "rxd_head",
3769 				CTLTYPE_UINT | CTLFLAG_RD, adapter,
3770 				E1000_RDH(rxr->me),
3771 				em_sysctl_reg_handler, "IU",
3772 				"Receive Descriptor Head");
3773 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "rxd_tail",
3774 				CTLTYPE_UINT | CTLFLAG_RD, adapter,
3775 				E1000_RDT(rxr->me),
3776 				em_sysctl_reg_handler, "IU",
3777 				"Receive Descriptor Tail");
3778 		SYSCTL_ADD_ULONG(ctx, queue_list, OID_AUTO, "rx_irq",
3779 				CTLFLAG_RD, &rxr->rx_irq,
3780 				"Queue MSI-X Receive Interrupts");
3781 	}
3782 
3783 	/* MAC stats get their own sub node */
3784 
3785 	stat_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, "mac_stats",
3786 				    CTLFLAG_RD, NULL, "Statistics");
3787 	stat_list = SYSCTL_CHILDREN(stat_node);
3788 
3789 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "excess_coll",
3790 			CTLFLAG_RD, &stats->ecol,
3791 			"Excessive collisions");
3792 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "single_coll",
3793 			CTLFLAG_RD, &stats->scc,
3794 			"Single collisions");
3795 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "multiple_coll",
3796 			CTLFLAG_RD, &stats->mcc,
3797 			"Multiple collisions");
3798 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "late_coll",
3799 			CTLFLAG_RD, &stats->latecol,
3800 			"Late collisions");
3801 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "collision_count",
3802 			CTLFLAG_RD, &stats->colc,
3803 			"Collision Count");
3804 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "symbol_errors",
3805 			CTLFLAG_RD, &adapter->stats.symerrs,
3806 			"Symbol Errors");
3807 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "sequence_errors",
3808 			CTLFLAG_RD, &adapter->stats.sec,
3809 			"Sequence Errors");
3810 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "defer_count",
3811 			CTLFLAG_RD, &adapter->stats.dc,
3812 			"Defer Count");
3813 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "missed_packets",
3814 			CTLFLAG_RD, &adapter->stats.mpc,
3815 			"Missed Packets");
3816 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_no_buff",
3817 			CTLFLAG_RD, &adapter->stats.rnbc,
3818 			"Receive No Buffers");
3819 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_undersize",
3820 			CTLFLAG_RD, &adapter->stats.ruc,
3821 			"Receive Undersize");
3822 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_fragmented",
3823 			CTLFLAG_RD, &adapter->stats.rfc,
3824 			"Fragmented Packets Received ");
3825 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_oversize",
3826 			CTLFLAG_RD, &adapter->stats.roc,
3827 			"Oversized Packets Received");
3828 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_jabber",
3829 			CTLFLAG_RD, &adapter->stats.rjc,
3830 			"Recevied Jabber");
3831 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_errs",
3832 			CTLFLAG_RD, &adapter->stats.rxerrc,
3833 			"Receive Errors");
3834 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "crc_errs",
3835 			CTLFLAG_RD, &adapter->stats.crcerrs,
3836 			"CRC errors");
3837 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "alignment_errs",
3838 			CTLFLAG_RD, &adapter->stats.algnerrc,
3839 			"Alignment Errors");
3840 	/* On 82575 these are collision counts */
3841 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "coll_ext_errs",
3842 			CTLFLAG_RD, &adapter->stats.cexterr,
3843 			"Collision/Carrier extension errors");
3844 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xon_recvd",
3845 			CTLFLAG_RD, &adapter->stats.xonrxc,
3846 			"XON Received");
3847 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xon_txd",
3848 			CTLFLAG_RD, &adapter->stats.xontxc,
3849 			"XON Transmitted");
3850 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xoff_recvd",
3851 			CTLFLAG_RD, &adapter->stats.xoffrxc,
3852 			"XOFF Received");
3853 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xoff_txd",
3854 			CTLFLAG_RD, &adapter->stats.xofftxc,
3855 			"XOFF Transmitted");
3856 
3857 	/* Packet Reception Stats */
3858 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_pkts_recvd",
3859 			CTLFLAG_RD, &adapter->stats.tpr,
3860 			"Total Packets Received ");
3861 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_pkts_recvd",
3862 			CTLFLAG_RD, &adapter->stats.gprc,
3863 			"Good Packets Received");
3864 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "bcast_pkts_recvd",
3865 			CTLFLAG_RD, &adapter->stats.bprc,
3866 			"Broadcast Packets Received");
3867 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "mcast_pkts_recvd",
3868 			CTLFLAG_RD, &adapter->stats.mprc,
3869 			"Multicast Packets Received");
3870 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_64",
3871 			CTLFLAG_RD, &adapter->stats.prc64,
3872 			"64 byte frames received ");
3873 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_65_127",
3874 			CTLFLAG_RD, &adapter->stats.prc127,
3875 			"65-127 byte frames received");
3876 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_128_255",
3877 			CTLFLAG_RD, &adapter->stats.prc255,
3878 			"128-255 byte frames received");
3879 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_256_511",
3880 			CTLFLAG_RD, &adapter->stats.prc511,
3881 			"256-511 byte frames received");
3882 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_512_1023",
3883 			CTLFLAG_RD, &adapter->stats.prc1023,
3884 			"512-1023 byte frames received");
3885 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_1024_1522",
3886 			CTLFLAG_RD, &adapter->stats.prc1522,
3887 			"1023-1522 byte frames received");
3888  	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_octets_recvd",
3889  			CTLFLAG_RD, &adapter->stats.gorc,
3890  			"Good Octets Received");
3891 
3892 	/* Packet Transmission Stats */
3893  	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_octets_txd",
3894  			CTLFLAG_RD, &adapter->stats.gotc,
3895  			"Good Octets Transmitted");
3896 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_pkts_txd",
3897 			CTLFLAG_RD, &adapter->stats.tpt,
3898 			"Total Packets Transmitted");
3899 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_pkts_txd",
3900 			CTLFLAG_RD, &adapter->stats.gptc,
3901 			"Good Packets Transmitted");
3902 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "bcast_pkts_txd",
3903 			CTLFLAG_RD, &adapter->stats.bptc,
3904 			"Broadcast Packets Transmitted");
3905 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "mcast_pkts_txd",
3906 			CTLFLAG_RD, &adapter->stats.mptc,
3907 			"Multicast Packets Transmitted");
3908 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_64",
3909 			CTLFLAG_RD, &adapter->stats.ptc64,
3910 			"64 byte frames transmitted ");
3911 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_65_127",
3912 			CTLFLAG_RD, &adapter->stats.ptc127,
3913 			"65-127 byte frames transmitted");
3914 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_128_255",
3915 			CTLFLAG_RD, &adapter->stats.ptc255,
3916 			"128-255 byte frames transmitted");
3917 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_256_511",
3918 			CTLFLAG_RD, &adapter->stats.ptc511,
3919 			"256-511 byte frames transmitted");
3920 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_512_1023",
3921 			CTLFLAG_RD, &adapter->stats.ptc1023,
3922 			"512-1023 byte frames transmitted");
3923 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_1024_1522",
3924 			CTLFLAG_RD, &adapter->stats.ptc1522,
3925 			"1024-1522 byte frames transmitted");
3926 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tso_txd",
3927 			CTLFLAG_RD, &adapter->stats.tsctc,
3928 			"TSO Contexts Transmitted");
3929 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tso_ctx_fail",
3930 			CTLFLAG_RD, &adapter->stats.tsctfc,
3931 			"TSO Contexts Failed");
3932 
3933 
3934 	/* Interrupt Stats */
3935 
3936 	int_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, "interrupts",
3937 				    CTLFLAG_RD, NULL, "Interrupt Statistics");
3938 	int_list = SYSCTL_CHILDREN(int_node);
3939 
3940 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "asserts",
3941 			CTLFLAG_RD, &adapter->stats.iac,
3942 			"Interrupt Assertion Count");
3943 
3944 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_pkt_timer",
3945 			CTLFLAG_RD, &adapter->stats.icrxptc,
3946 			"Interrupt Cause Rx Pkt Timer Expire Count");
3947 
3948 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_abs_timer",
3949 			CTLFLAG_RD, &adapter->stats.icrxatc,
3950 			"Interrupt Cause Rx Abs Timer Expire Count");
3951 
3952 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_pkt_timer",
3953 			CTLFLAG_RD, &adapter->stats.ictxptc,
3954 			"Interrupt Cause Tx Pkt Timer Expire Count");
3955 
3956 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_abs_timer",
3957 			CTLFLAG_RD, &adapter->stats.ictxatc,
3958 			"Interrupt Cause Tx Abs Timer Expire Count");
3959 
3960 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_queue_empty",
3961 			CTLFLAG_RD, &adapter->stats.ictxqec,
3962 			"Interrupt Cause Tx Queue Empty Count");
3963 
3964 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_queue_min_thresh",
3965 			CTLFLAG_RD, &adapter->stats.ictxqmtc,
3966 			"Interrupt Cause Tx Queue Min Thresh Count");
3967 
3968 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_desc_min_thresh",
3969 			CTLFLAG_RD, &adapter->stats.icrxdmtc,
3970 			"Interrupt Cause Rx Desc Min Thresh Count");
3971 
3972 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_overrun",
3973 			CTLFLAG_RD, &adapter->stats.icrxoc,
3974 			"Interrupt Cause Receiver Overrun Count");
3975 }
3976 
3977 /**********************************************************************
3978  *
3979  *  This routine provides a way to dump out the adapter eeprom,
3980  *  often a useful debug/service tool. This only dumps the first
3981  *  32 words, stuff that matters is in that extent.
3982  *
3983  **********************************************************************/
3984 static int
3985 em_sysctl_nvm_info(SYSCTL_HANDLER_ARGS)
3986 {
3987 	struct adapter *adapter = (struct adapter *)arg1;
3988 	int error;
3989 	int result;
3990 
3991 	result = -1;
3992 	error = sysctl_handle_int(oidp, &result, 0, req);
3993 
3994 	if (error || !req->newptr)
3995 		return (error);
3996 
3997 	/*
3998 	 * This value will cause a hex dump of the
3999 	 * first 32 16-bit words of the EEPROM to
4000 	 * the screen.
4001 	 */
4002 	if (result == 1)
4003 		em_print_nvm_info(adapter);
4004 
4005 	return (error);
4006 }
4007 
4008 static void
4009 em_print_nvm_info(struct adapter *adapter)
4010 {
4011 	u16	eeprom_data;
4012 	int	i, j, row = 0;
4013 
4014 	/* Its a bit crude, but it gets the job done */
4015 	printf("\nInterface EEPROM Dump:\n");
4016 	printf("Offset\n0x0000  ");
4017 	for (i = 0, j = 0; i < 32; i++, j++) {
4018 		if (j == 8) { /* Make the offset block */
4019 			j = 0; ++row;
4020 			printf("\n0x00%x0  ",row);
4021 		}
4022 		e1000_read_nvm(&adapter->hw, i, 1, &eeprom_data);
4023 		printf("%04x ", eeprom_data);
4024 	}
4025 	printf("\n");
4026 }
4027 
4028 static int
4029 em_sysctl_int_delay(SYSCTL_HANDLER_ARGS)
4030 {
4031 	struct em_int_delay_info *info;
4032 	struct adapter *adapter;
4033 	u32 regval;
4034 	int error, usecs, ticks;
4035 
4036 	info = (struct em_int_delay_info *)arg1;
4037 	usecs = info->value;
4038 	error = sysctl_handle_int(oidp, &usecs, 0, req);
4039 	if (error != 0 || req->newptr == NULL)
4040 		return (error);
4041 	if (usecs < 0 || usecs > EM_TICKS_TO_USECS(65535))
4042 		return (EINVAL);
4043 	info->value = usecs;
4044 	ticks = EM_USECS_TO_TICKS(usecs);
4045 	if (info->offset == E1000_ITR)	/* units are 256ns here */
4046 		ticks *= 4;
4047 
4048 	adapter = info->adapter;
4049 
4050 	regval = E1000_READ_OFFSET(&adapter->hw, info->offset);
4051 	regval = (regval & ~0xffff) | (ticks & 0xffff);
4052 	/* Handle a few special cases. */
4053 	switch (info->offset) {
4054 	case E1000_RDTR:
4055 		break;
4056 	case E1000_TIDV:
4057 		if (ticks == 0) {
4058 			adapter->txd_cmd &= ~E1000_TXD_CMD_IDE;
4059 			/* Don't write 0 into the TIDV register. */
4060 			regval++;
4061 		} else
4062 			adapter->txd_cmd |= E1000_TXD_CMD_IDE;
4063 		break;
4064 	}
4065 	E1000_WRITE_OFFSET(&adapter->hw, info->offset, regval);
4066 	return (0);
4067 }
4068 
4069 static void
4070 em_add_int_delay_sysctl(struct adapter *adapter, const char *name,
4071 	const char *description, struct em_int_delay_info *info,
4072 	int offset, int value)
4073 {
4074 	info->adapter = adapter;
4075 	info->offset = offset;
4076 	info->value = value;
4077 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(adapter->dev),
4078 	    SYSCTL_CHILDREN(device_get_sysctl_tree(adapter->dev)),
4079 	    OID_AUTO, name, CTLTYPE_INT|CTLFLAG_RW,
4080 	    info, 0, em_sysctl_int_delay, "I", description);
4081 }
4082 
4083 static void
4084 em_set_sysctl_value(struct adapter *adapter, const char *name,
4085 	const char *description, int *limit, int value)
4086 {
4087 	*limit = value;
4088 	SYSCTL_ADD_INT(device_get_sysctl_ctx(adapter->dev),
4089 	    SYSCTL_CHILDREN(device_get_sysctl_tree(adapter->dev)),
4090 	    OID_AUTO, name, CTLFLAG_RW, limit, value, description);
4091 }
4092 
4093 
4094 /*
4095 ** Set flow control using sysctl:
4096 ** Flow control values:
4097 **      0 - off
4098 **      1 - rx pause
4099 **      2 - tx pause
4100 **      3 - full
4101 */
4102 static int
4103 em_set_flowcntl(SYSCTL_HANDLER_ARGS)
4104 {
4105         int		error;
4106 	static int	input = 3; /* default is full */
4107         struct adapter	*adapter = (struct adapter *) arg1;
4108 
4109         error = sysctl_handle_int(oidp, &input, 0, req);
4110 
4111         if ((error) || (req->newptr == NULL))
4112                 return (error);
4113 
4114 	if (input == adapter->fc) /* no change? */
4115 		return (error);
4116 
4117         switch (input) {
4118                 case e1000_fc_rx_pause:
4119                 case e1000_fc_tx_pause:
4120                 case e1000_fc_full:
4121                 case e1000_fc_none:
4122                         adapter->hw.fc.requested_mode = input;
4123 			adapter->fc = input;
4124                         break;
4125                 default:
4126 			/* Do nothing */
4127 			return (error);
4128         }
4129 
4130         adapter->hw.fc.current_mode = adapter->hw.fc.requested_mode;
4131         e1000_force_mac_fc(&adapter->hw);
4132         return (error);
4133 }
4134 
4135 /*
4136 ** Manage Energy Efficient Ethernet:
4137 ** Control values:
4138 **     0/1 - enabled/disabled
4139 */
4140 static int
4141 em_sysctl_eee(SYSCTL_HANDLER_ARGS)
4142 {
4143        struct adapter *adapter = (struct adapter *) arg1;
4144        int             error, value;
4145 
4146        value = adapter->hw.dev_spec.ich8lan.eee_disable;
4147        error = sysctl_handle_int(oidp, &value, 0, req);
4148        if (error || req->newptr == NULL)
4149                return (error);
4150        adapter->hw.dev_spec.ich8lan.eee_disable = (value != 0);
4151        em_if_init(adapter->ctx);
4152 
4153        return (0);
4154 }
4155 
4156 static int
4157 em_sysctl_debug_info(SYSCTL_HANDLER_ARGS)
4158 {
4159 	struct adapter *adapter;
4160 	int error;
4161 	int result;
4162 
4163 	result = -1;
4164 	error = sysctl_handle_int(oidp, &result, 0, req);
4165 
4166 	if (error || !req->newptr)
4167 		return (error);
4168 
4169 	if (result == 1) {
4170 		adapter = (struct adapter *)arg1;
4171 		em_print_debug_info(adapter);
4172         }
4173 
4174 	return (error);
4175 }
4176 
4177 /*
4178 ** This routine is meant to be fluid, add whatever is
4179 ** needed for debugging a problem.  -jfv
4180 */
4181 static void
4182 em_print_debug_info(struct adapter *adapter)
4183 {
4184 	device_t dev = adapter->dev;
4185 	struct tx_ring *txr = &adapter->tx_queues->txr;
4186 	struct rx_ring *rxr = &adapter->rx_queues->rxr;
4187 
4188 	if (if_getdrvflags(adapter->ifp) & IFF_DRV_RUNNING)
4189 		printf("Interface is RUNNING ");
4190 	else
4191 		printf("Interface is NOT RUNNING\n");
4192 
4193 	if (if_getdrvflags(adapter->ifp) & IFF_DRV_OACTIVE)
4194 		printf("and INACTIVE\n");
4195 	else
4196 		printf("and ACTIVE\n");
4197 
4198 	for (int i = 0; i < adapter->tx_num_queues; i++, txr++) {
4199 		device_printf(dev, "TX Queue %d ------\n", i);
4200 		device_printf(dev, "hw tdh = %d, hw tdt = %d\n",
4201 	    		E1000_READ_REG(&adapter->hw, E1000_TDH(i)),
4202 	    		E1000_READ_REG(&adapter->hw, E1000_TDT(i)));
4203 
4204 	}
4205 	for (int j=0; j < adapter->rx_num_queues; j++, rxr++) {
4206 		device_printf(dev, "RX Queue %d ------\n", j);
4207 		device_printf(dev, "hw rdh = %d, hw rdt = %d\n",
4208 	    		E1000_READ_REG(&adapter->hw, E1000_RDH(j)),
4209 	    		E1000_READ_REG(&adapter->hw, E1000_RDT(j)));
4210 	}
4211 }
4212 
4213 
4214 /*
4215  * 82574 only:
4216  * Write a new value to the EEPROM increasing the number of MSIX
4217  * vectors from 3 to 5, for proper multiqueue support.
4218  */
4219 static void
4220 em_enable_vectors_82574(if_ctx_t ctx)
4221 {
4222 	struct adapter *adapter = iflib_get_softc(ctx);
4223 	struct e1000_hw *hw = &adapter->hw;
4224 	device_t dev = iflib_get_dev(ctx);
4225 	u16 edata;
4226 
4227 	e1000_read_nvm(hw, EM_NVM_PCIE_CTRL, 1, &edata);
4228 	printf("Current cap: %#06x\n", edata);
4229 	if (((edata & EM_NVM_MSIX_N_MASK) >> EM_NVM_MSIX_N_SHIFT) != 4) {
4230 		device_printf(dev, "Writing to eeprom: increasing "
4231 		    "reported MSIX vectors from 3 to 5...\n");
4232 		edata &= ~(EM_NVM_MSIX_N_MASK);
4233 		edata |= 4 << EM_NVM_MSIX_N_SHIFT;
4234 		e1000_write_nvm(hw, EM_NVM_PCIE_CTRL, 1, &edata);
4235 		e1000_update_nvm_checksum(hw);
4236 		device_printf(dev, "Writing to eeprom: done\n");
4237 	}
4238 }
4239 
4240 
4241 #ifdef DDB
4242 DB_COMMAND(em_reset_dev, em_ddb_reset_dev)
4243 {
4244 	devclass_t	dc;
4245 	int max_em;
4246 
4247 	dc = devclass_find("em");
4248 	max_em = devclass_get_maxunit(dc);
4249 
4250 	for (int index = 0; index < (max_em - 1); index++) {
4251 		device_t dev;
4252 		dev = devclass_get_device(dc, index);
4253 		if (device_get_driver(dev) == &em_driver) {
4254 			struct adapter *adapter = device_get_softc(dev);
4255 			em_if_init(adapter->ctx);
4256 		}
4257 	}
4258 }
4259 DB_COMMAND(em_dump_queue, em_ddb_dump_queue)
4260 {
4261 	devclass_t	dc;
4262 	int max_em;
4263 
4264 	dc = devclass_find("em");
4265 	max_em = devclass_get_maxunit(dc);
4266 
4267 	for (int index = 0; index < (max_em - 1); index++) {
4268 		device_t dev;
4269 		dev = devclass_get_device(dc, index);
4270 		if (device_get_driver(dev) == &em_driver)
4271 			em_print_debug_info(device_get_softc(dev));
4272 	}
4273 
4274 }
4275 #endif
4276