]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/blob - sys/dev/e1000/if_em.c
sqlite3: Vendor import of sqlite3 3.38.5
[FreeBSD/FreeBSD.git] / sys / dev / e1000 / if_em.c
1 /*-
2  * SPDX-License-Identifier: BSD-2-Clause
3  *
4  * Copyright (c) 2016 Nicole Graziano <nicole@nextbsd.org>
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  *
16  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
17  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
20  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
26  * SUCH DAMAGE.
27  */
28
29 /* $FreeBSD$ */
30 #include "if_em.h"
31 #include <sys/sbuf.h>
32 #include <machine/_inttypes.h>
33
34 #define em_mac_min e1000_82571
35 #define igb_mac_min e1000_82575
36
37 /*********************************************************************
38  *  Driver version:
39  *********************************************************************/
40 char em_driver_version[] = "7.6.1-k";
41
42 /*********************************************************************
43  *  PCI Device ID Table
44  *
45  *  Used by probe to select devices to load on
46  *  Last field stores an index into e1000_strings
47  *  Last entry must be all 0s
48  *
49  *  { Vendor ID, Device ID, SubVendor ID, SubDevice ID, String Index }
50  *********************************************************************/
51
52 static pci_vendor_info_t em_vendor_info_array[] =
53 {
54         /* Intel(R) - lem-class legacy devices */
55         PVID(0x8086, E1000_DEV_ID_82540EM, "Intel(R) Legacy PRO/1000 MT 82540EM"),
56         PVID(0x8086, E1000_DEV_ID_82540EM_LOM, "Intel(R) Legacy PRO/1000 MT 82540EM (LOM)"),
57         PVID(0x8086, E1000_DEV_ID_82540EP, "Intel(R) Legacy PRO/1000 MT 82540EP"),
58         PVID(0x8086, E1000_DEV_ID_82540EP_LOM, "Intel(R) Legacy PRO/1000 MT 82540EP (LOM)"),
59         PVID(0x8086, E1000_DEV_ID_82540EP_LP, "Intel(R) Legacy PRO/1000 MT 82540EP (Mobile)"),
60
61         PVID(0x8086, E1000_DEV_ID_82541EI, "Intel(R) Legacy PRO/1000 MT 82541EI (Copper)"),
62         PVID(0x8086, E1000_DEV_ID_82541ER, "Intel(R) Legacy PRO/1000 82541ER"),
63         PVID(0x8086, E1000_DEV_ID_82541ER_LOM, "Intel(R) Legacy PRO/1000 MT 82541ER"),
64         PVID(0x8086, E1000_DEV_ID_82541EI_MOBILE, "Intel(R) Legacy PRO/1000 MT 82541EI (Mobile)"),
65         PVID(0x8086, E1000_DEV_ID_82541GI, "Intel(R) Legacy PRO/1000 MT 82541GI"),
66         PVID(0x8086, E1000_DEV_ID_82541GI_LF, "Intel(R) Legacy PRO/1000 GT 82541PI"),
67         PVID(0x8086, E1000_DEV_ID_82541GI_MOBILE, "Intel(R) Legacy PRO/1000 MT 82541GI (Mobile)"),
68
69         PVID(0x8086, E1000_DEV_ID_82542, "Intel(R) Legacy PRO/1000 82542 (Fiber)"),
70
71         PVID(0x8086, E1000_DEV_ID_82543GC_FIBER, "Intel(R) Legacy PRO/1000 F 82543GC (Fiber)"),
72         PVID(0x8086, E1000_DEV_ID_82543GC_COPPER, "Intel(R) Legacy PRO/1000 T 82543GC (Copper)"),
73
74         PVID(0x8086, E1000_DEV_ID_82544EI_COPPER, "Intel(R) Legacy PRO/1000 XT 82544EI (Copper)"),
75         PVID(0x8086, E1000_DEV_ID_82544EI_FIBER, "Intel(R) Legacy PRO/1000 XF 82544EI (Fiber)"),
76         PVID(0x8086, E1000_DEV_ID_82544GC_COPPER, "Intel(R) Legacy PRO/1000 T 82544GC (Copper)"),
77         PVID(0x8086, E1000_DEV_ID_82544GC_LOM, "Intel(R) Legacy PRO/1000 XT 82544GC (LOM)"),
78
79         PVID(0x8086, E1000_DEV_ID_82545EM_COPPER, "Intel(R) Legacy PRO/1000 MT 82545EM (Copper)"),
80         PVID(0x8086, E1000_DEV_ID_82545EM_FIBER, "Intel(R) Legacy PRO/1000 MF 82545EM (Fiber)"),
81         PVID(0x8086, E1000_DEV_ID_82545GM_COPPER, "Intel(R) Legacy PRO/1000 MT 82545GM (Copper)"),
82         PVID(0x8086, E1000_DEV_ID_82545GM_FIBER, "Intel(R) Legacy PRO/1000 MF 82545GM (Fiber)"),
83         PVID(0x8086, E1000_DEV_ID_82545GM_SERDES, "Intel(R) Legacy PRO/1000 MB 82545GM (SERDES)"),
84
85         PVID(0x8086, E1000_DEV_ID_82546EB_COPPER, "Intel(R) Legacy PRO/1000 MT 82546EB (Copper)"),
86         PVID(0x8086, E1000_DEV_ID_82546EB_FIBER, "Intel(R) Legacy PRO/1000 MF 82546EB (Fiber)"),
87         PVID(0x8086, E1000_DEV_ID_82546EB_QUAD_COPPER, "Intel(R) Legacy PRO/1000 MT 82546EB (Quad Copper"),
88         PVID(0x8086, E1000_DEV_ID_82546GB_COPPER, "Intel(R) Legacy PRO/1000 MT 82546GB (Copper)"),
89         PVID(0x8086, E1000_DEV_ID_82546GB_FIBER, "Intel(R) Legacy PRO/1000 MF 82546GB (Fiber)"),
90         PVID(0x8086, E1000_DEV_ID_82546GB_SERDES, "Intel(R) Legacy PRO/1000 MB 82546GB (SERDES)"),
91         PVID(0x8086, E1000_DEV_ID_82546GB_PCIE, "Intel(R) Legacy PRO/1000 P 82546GB (PCIe)"),
92         PVID(0x8086, E1000_DEV_ID_82546GB_QUAD_COPPER, "Intel(R) Legacy PRO/1000 GT 82546GB (Quad Copper)"),
93         PVID(0x8086, E1000_DEV_ID_82546GB_QUAD_COPPER_KSP3, "Intel(R) Legacy PRO/1000 GT 82546GB (Quad Copper)"),
94
95         PVID(0x8086, E1000_DEV_ID_82547EI, "Intel(R) Legacy PRO/1000 CT 82547EI"),
96         PVID(0x8086, E1000_DEV_ID_82547EI_MOBILE, "Intel(R) Legacy PRO/1000 CT 82547EI (Mobile)"),
97         PVID(0x8086, E1000_DEV_ID_82547GI, "Intel(R) Legacy PRO/1000 CT 82547GI"),
98
99         /* Intel(R) - em-class devices */
100         PVID(0x8086, E1000_DEV_ID_82571EB_COPPER, "Intel(R) PRO/1000 PT 82571EB/82571GB (Copper)"),
101         PVID(0x8086, E1000_DEV_ID_82571EB_FIBER, "Intel(R) PRO/1000 PF 82571EB/82571GB (Fiber)"),
102         PVID(0x8086, E1000_DEV_ID_82571EB_SERDES, "Intel(R) PRO/1000 PB 82571EB (SERDES)"),
103         PVID(0x8086, E1000_DEV_ID_82571EB_SERDES_DUAL, "Intel(R) PRO/1000 82571EB (Dual Mezzanine)"),
104         PVID(0x8086, E1000_DEV_ID_82571EB_SERDES_QUAD, "Intel(R) PRO/1000 82571EB (Quad Mezzanine)"),
105         PVID(0x8086, E1000_DEV_ID_82571EB_QUAD_COPPER, "Intel(R) PRO/1000 PT 82571EB/82571GB (Quad Copper)"),
106         PVID(0x8086, E1000_DEV_ID_82571EB_QUAD_COPPER_LP, "Intel(R) PRO/1000 PT 82571EB/82571GB (Quad Copper)"),
107         PVID(0x8086, E1000_DEV_ID_82571EB_QUAD_FIBER, "Intel(R) PRO/1000 PF 82571EB (Quad Fiber)"),
108         PVID(0x8086, E1000_DEV_ID_82571PT_QUAD_COPPER, "Intel(R) PRO/1000 PT 82571PT (Quad Copper)"),
109         PVID(0x8086, E1000_DEV_ID_82572EI, "Intel(R) PRO/1000 PT 82572EI (Copper)"),
110         PVID(0x8086, E1000_DEV_ID_82572EI_COPPER, "Intel(R) PRO/1000 PT 82572EI (Copper)"),
111         PVID(0x8086, E1000_DEV_ID_82572EI_FIBER, "Intel(R) PRO/1000 PF 82572EI (Fiber)"),
112         PVID(0x8086, E1000_DEV_ID_82572EI_SERDES, "Intel(R) PRO/1000 82572EI (SERDES)"),
113         PVID(0x8086, E1000_DEV_ID_82573E, "Intel(R) PRO/1000 82573E (Copper)"),
114         PVID(0x8086, E1000_DEV_ID_82573E_IAMT, "Intel(R) PRO/1000 82573E AMT (Copper)"),
115         PVID(0x8086, E1000_DEV_ID_82573L, "Intel(R) PRO/1000 82573L"),
116         PVID(0x8086, E1000_DEV_ID_82583V, "Intel(R) 82583V"),
117         PVID(0x8086, E1000_DEV_ID_80003ES2LAN_COPPER_SPT, "Intel(R) 80003ES2LAN (Copper)"),
118         PVID(0x8086, E1000_DEV_ID_80003ES2LAN_SERDES_SPT, "Intel(R) 80003ES2LAN (SERDES)"),
119         PVID(0x8086, E1000_DEV_ID_80003ES2LAN_COPPER_DPT, "Intel(R) 80003ES2LAN (Dual Copper)"),
120         PVID(0x8086, E1000_DEV_ID_80003ES2LAN_SERDES_DPT, "Intel(R) 80003ES2LAN (Dual SERDES)"),
121         PVID(0x8086, E1000_DEV_ID_ICH8_IGP_M_AMT, "Intel(R) 82566MM ICH8 AMT (Mobile)"),
122         PVID(0x8086, E1000_DEV_ID_ICH8_IGP_AMT, "Intel(R) 82566DM ICH8 AMT"),
123         PVID(0x8086, E1000_DEV_ID_ICH8_IGP_C, "Intel(R) 82566DC ICH8"),
124         PVID(0x8086, E1000_DEV_ID_ICH8_IFE, "Intel(R) 82562V ICH8"),
125         PVID(0x8086, E1000_DEV_ID_ICH8_IFE_GT, "Intel(R) 82562GT ICH8"),
126         PVID(0x8086, E1000_DEV_ID_ICH8_IFE_G, "Intel(R) 82562G ICH8"),
127         PVID(0x8086, E1000_DEV_ID_ICH8_IGP_M, "Intel(R) 82566MC ICH8"),
128         PVID(0x8086, E1000_DEV_ID_ICH8_82567V_3, "Intel(R) 82567V-3 ICH8"),
129         PVID(0x8086, E1000_DEV_ID_ICH9_IGP_M_AMT, "Intel(R) 82567LM ICH9 AMT"),
130         PVID(0x8086, E1000_DEV_ID_ICH9_IGP_AMT, "Intel(R) 82566DM-2 ICH9 AMT"),
131         PVID(0x8086, E1000_DEV_ID_ICH9_IGP_C, "Intel(R) 82566DC-2 ICH9"),
132         PVID(0x8086, E1000_DEV_ID_ICH9_IGP_M, "Intel(R) 82567LF ICH9"),
133         PVID(0x8086, E1000_DEV_ID_ICH9_IGP_M_V, "Intel(R) 82567V ICH9"),
134         PVID(0x8086, E1000_DEV_ID_ICH9_IFE, "Intel(R) 82562V-2 ICH9"),
135         PVID(0x8086, E1000_DEV_ID_ICH9_IFE_GT, "Intel(R) 82562GT-2 ICH9"),
136         PVID(0x8086, E1000_DEV_ID_ICH9_IFE_G, "Intel(R) 82562G-2 ICH9"),
137         PVID(0x8086, E1000_DEV_ID_ICH9_BM, "Intel(R) 82567LM-4 ICH9"),
138         PVID(0x8086, E1000_DEV_ID_82574L, "Intel(R) Gigabit CT 82574L"),
139         PVID(0x8086, E1000_DEV_ID_82574LA, "Intel(R) 82574L-Apple"),
140         PVID(0x8086, E1000_DEV_ID_ICH10_R_BM_LM, "Intel(R) 82567LM-2 ICH10"),
141         PVID(0x8086, E1000_DEV_ID_ICH10_R_BM_LF, "Intel(R) 82567LF-2 ICH10"),
142         PVID(0x8086, E1000_DEV_ID_ICH10_R_BM_V, "Intel(R) 82567V-2 ICH10"),
143         PVID(0x8086, E1000_DEV_ID_ICH10_D_BM_LM, "Intel(R) 82567LM-3 ICH10"),
144         PVID(0x8086, E1000_DEV_ID_ICH10_D_BM_LF, "Intel(R) 82567LF-3 ICH10"),
145         PVID(0x8086, E1000_DEV_ID_ICH10_D_BM_V, "Intel(R) 82567V-4 ICH10"),
146         PVID(0x8086, E1000_DEV_ID_PCH_M_HV_LM, "Intel(R) 82577LM"),
147         PVID(0x8086, E1000_DEV_ID_PCH_M_HV_LC, "Intel(R) 82577LC"),
148         PVID(0x8086, E1000_DEV_ID_PCH_D_HV_DM, "Intel(R) 82578DM"),
149         PVID(0x8086, E1000_DEV_ID_PCH_D_HV_DC, "Intel(R) 82578DC"),
150         PVID(0x8086, E1000_DEV_ID_PCH2_LV_LM, "Intel(R) 82579LM"),
151         PVID(0x8086, E1000_DEV_ID_PCH2_LV_V, "Intel(R) 82579V"),
152         PVID(0x8086, E1000_DEV_ID_PCH_LPT_I217_LM, "Intel(R) I217-LM LPT"),
153         PVID(0x8086, E1000_DEV_ID_PCH_LPT_I217_V, "Intel(R) I217-V LPT"),
154         PVID(0x8086, E1000_DEV_ID_PCH_LPTLP_I218_LM, "Intel(R) I218-LM LPTLP"),
155         PVID(0x8086, E1000_DEV_ID_PCH_LPTLP_I218_V, "Intel(R) I218-V LPTLP"),
156         PVID(0x8086, E1000_DEV_ID_PCH_I218_LM2, "Intel(R) I218-LM (2)"),
157         PVID(0x8086, E1000_DEV_ID_PCH_I218_V2, "Intel(R) I218-V (2)"),
158         PVID(0x8086, E1000_DEV_ID_PCH_I218_LM3, "Intel(R) I218-LM (3)"),
159         PVID(0x8086, E1000_DEV_ID_PCH_I218_V3, "Intel(R) I218-V (3)"),
160         PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_LM, "Intel(R) I219-LM SPT"),
161         PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_V, "Intel(R) I219-V SPT"),
162         PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_LM2, "Intel(R) I219-LM SPT-H(2)"),
163         PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_V2, "Intel(R) I219-V SPT-H(2)"),
164         PVID(0x8086, E1000_DEV_ID_PCH_LBG_I219_LM3, "Intel(R) I219-LM LBG(3)"),
165         PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_LM4, "Intel(R) I219-LM SPT(4)"),
166         PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_V4, "Intel(R) I219-V SPT(4)"),
167         PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_LM5, "Intel(R) I219-LM SPT(5)"),
168         PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_V5, "Intel(R) I219-V SPT(5)"),
169         PVID(0x8086, E1000_DEV_ID_PCH_CNP_I219_LM6, "Intel(R) I219-LM CNP(6)"),
170         PVID(0x8086, E1000_DEV_ID_PCH_CNP_I219_V6, "Intel(R) I219-V CNP(6)"),
171         PVID(0x8086, E1000_DEV_ID_PCH_CNP_I219_LM7, "Intel(R) I219-LM CNP(7)"),
172         PVID(0x8086, E1000_DEV_ID_PCH_CNP_I219_V7, "Intel(R) I219-V CNP(7)"),
173         PVID(0x8086, E1000_DEV_ID_PCH_ICP_I219_LM8, "Intel(R) I219-LM ICP(8)"),
174         PVID(0x8086, E1000_DEV_ID_PCH_ICP_I219_V8, "Intel(R) I219-V ICP(8)"),
175         PVID(0x8086, E1000_DEV_ID_PCH_ICP_I219_LM9, "Intel(R) I219-LM ICP(9)"),
176         PVID(0x8086, E1000_DEV_ID_PCH_ICP_I219_V9, "Intel(R) I219-V ICP(9)"),
177         PVID(0x8086, E1000_DEV_ID_PCH_CMP_I219_LM10, "Intel(R) I219-LM CMP(10)"),
178         PVID(0x8086, E1000_DEV_ID_PCH_CMP_I219_V10, "Intel(R) I219-V CMP(10)"),
179         PVID(0x8086, E1000_DEV_ID_PCH_CMP_I219_LM11, "Intel(R) I219-LM CMP(11)"),
180         PVID(0x8086, E1000_DEV_ID_PCH_CMP_I219_V11, "Intel(R) I219-V CMP(11)"),
181         PVID(0x8086, E1000_DEV_ID_PCH_CMP_I219_LM12, "Intel(R) I219-LM CMP(12)"),
182         PVID(0x8086, E1000_DEV_ID_PCH_CMP_I219_V12, "Intel(R) I219-V CMP(12)"),
183         PVID(0x8086, E1000_DEV_ID_PCH_TGP_I219_LM13, "Intel(R) I219-LM TGP(13)"),
184         PVID(0x8086, E1000_DEV_ID_PCH_TGP_I219_V13, "Intel(R) I219-V TGP(13)"),
185         PVID(0x8086, E1000_DEV_ID_PCH_TGP_I219_LM14, "Intel(R) I219-LM TGP(14)"),
186         PVID(0x8086, E1000_DEV_ID_PCH_TGP_I219_V14, "Intel(R) I219-V GTP(14)"),
187         PVID(0x8086, E1000_DEV_ID_PCH_TGP_I219_LM15, "Intel(R) I219-LM TGP(15)"),
188         PVID(0x8086, E1000_DEV_ID_PCH_TGP_I219_V15, "Intel(R) I219-V TGP(15)"),
189         PVID(0x8086, E1000_DEV_ID_PCH_ADL_I219_LM16, "Intel(R) I219-LM ADL(16)"),
190         PVID(0x8086, E1000_DEV_ID_PCH_ADL_I219_V16, "Intel(R) I219-V ADL(16)"),
191         PVID(0x8086, E1000_DEV_ID_PCH_ADL_I219_LM17, "Intel(R) I219-LM ADL(17)"),
192         PVID(0x8086, E1000_DEV_ID_PCH_ADL_I219_V17, "Intel(R) I219-V ADL(17)"),
193         PVID(0x8086, E1000_DEV_ID_PCH_MTP_I219_LM18, "Intel(R) I219-LM MTP(18)"),
194         PVID(0x8086, E1000_DEV_ID_PCH_MTP_I219_V18, "Intel(R) I219-V MTP(18)"),
195         PVID(0x8086, E1000_DEV_ID_PCH_MTP_I219_LM19, "Intel(R) I219-LM MTP(19)"),
196         PVID(0x8086, E1000_DEV_ID_PCH_MTP_I219_V19, "Intel(R) I219-V MTP(19)"),
197         /* required last entry */
198         PVID_END
199 };
200
201 static pci_vendor_info_t igb_vendor_info_array[] =
202 {
203         /* Intel(R) - igb-class devices */
204         PVID(0x8086, E1000_DEV_ID_82575EB_COPPER, "Intel(R) PRO/1000 82575EB (Copper)"),
205         PVID(0x8086, E1000_DEV_ID_82575EB_FIBER_SERDES, "Intel(R) PRO/1000 82575EB (SERDES)"),
206         PVID(0x8086, E1000_DEV_ID_82575GB_QUAD_COPPER, "Intel(R) PRO/1000 VT 82575GB (Quad Copper)"),
207         PVID(0x8086, E1000_DEV_ID_82576, "Intel(R) PRO/1000 82576"),
208         PVID(0x8086, E1000_DEV_ID_82576_NS, "Intel(R) PRO/1000 82576NS"),
209         PVID(0x8086, E1000_DEV_ID_82576_NS_SERDES, "Intel(R) PRO/1000 82576NS (SERDES)"),
210         PVID(0x8086, E1000_DEV_ID_82576_FIBER, "Intel(R) PRO/1000 EF 82576 (Dual Fiber)"),
211         PVID(0x8086, E1000_DEV_ID_82576_SERDES, "Intel(R) PRO/1000 82576 (Dual SERDES)"),
212         PVID(0x8086, E1000_DEV_ID_82576_SERDES_QUAD, "Intel(R) PRO/1000 ET 82576 (Quad SERDES)"),
213         PVID(0x8086, E1000_DEV_ID_82576_QUAD_COPPER, "Intel(R) PRO/1000 ET 82576 (Quad Copper)"),
214         PVID(0x8086, E1000_DEV_ID_82576_QUAD_COPPER_ET2, "Intel(R) PRO/1000 ET(2) 82576 (Quad Copper)"),
215         PVID(0x8086, E1000_DEV_ID_82576_VF, "Intel(R) PRO/1000 82576 Virtual Function"),
216         PVID(0x8086, E1000_DEV_ID_82580_COPPER, "Intel(R) I340 82580 (Copper)"),
217         PVID(0x8086, E1000_DEV_ID_82580_FIBER, "Intel(R) I340 82580 (Fiber)"),
218         PVID(0x8086, E1000_DEV_ID_82580_SERDES, "Intel(R) I340 82580 (SERDES)"),
219         PVID(0x8086, E1000_DEV_ID_82580_SGMII, "Intel(R) I340 82580 (SGMII)"),
220         PVID(0x8086, E1000_DEV_ID_82580_COPPER_DUAL, "Intel(R) I340-T2 82580 (Dual Copper)"),
221         PVID(0x8086, E1000_DEV_ID_82580_QUAD_FIBER, "Intel(R) I340-F4 82580 (Quad Fiber)"),
222         PVID(0x8086, E1000_DEV_ID_DH89XXCC_SERDES, "Intel(R) DH89XXCC (SERDES)"),
223         PVID(0x8086, E1000_DEV_ID_DH89XXCC_SGMII, "Intel(R) I347-AT4 DH89XXCC"),
224         PVID(0x8086, E1000_DEV_ID_DH89XXCC_SFP, "Intel(R) DH89XXCC (SFP)"),
225         PVID(0x8086, E1000_DEV_ID_DH89XXCC_BACKPLANE, "Intel(R) DH89XXCC (Backplane)"),
226         PVID(0x8086, E1000_DEV_ID_I350_COPPER, "Intel(R) I350 (Copper)"),
227         PVID(0x8086, E1000_DEV_ID_I350_FIBER, "Intel(R) I350 (Fiber)"),
228         PVID(0x8086, E1000_DEV_ID_I350_SERDES, "Intel(R) I350 (SERDES)"),
229         PVID(0x8086, E1000_DEV_ID_I350_SGMII, "Intel(R) I350 (SGMII)"),
230         PVID(0x8086, E1000_DEV_ID_I350_VF, "Intel(R) I350 Virtual Function"),
231         PVID(0x8086, E1000_DEV_ID_I210_COPPER, "Intel(R) I210 (Copper)"),
232         PVID(0x8086, E1000_DEV_ID_I210_COPPER_IT, "Intel(R) I210 IT (Copper)"),
233         PVID(0x8086, E1000_DEV_ID_I210_COPPER_OEM1, "Intel(R) I210 (OEM)"),
234         PVID(0x8086, E1000_DEV_ID_I210_COPPER_FLASHLESS, "Intel(R) I210 Flashless (Copper)"),
235         PVID(0x8086, E1000_DEV_ID_I210_SERDES_FLASHLESS, "Intel(R) I210 Flashless (SERDES)"),
236         PVID(0x8086, E1000_DEV_ID_I210_SGMII_FLASHLESS, "Intel(R) I210 Flashless (SGMII)"),
237         PVID(0x8086, E1000_DEV_ID_I210_FIBER, "Intel(R) I210 (Fiber)"),
238         PVID(0x8086, E1000_DEV_ID_I210_SERDES, "Intel(R) I210 (SERDES)"),
239         PVID(0x8086, E1000_DEV_ID_I210_SGMII, "Intel(R) I210 (SGMII)"),
240         PVID(0x8086, E1000_DEV_ID_I211_COPPER, "Intel(R) I211 (Copper)"),
241         PVID(0x8086, E1000_DEV_ID_I354_BACKPLANE_1GBPS, "Intel(R) I354 (1.0 GbE Backplane)"),
242         PVID(0x8086, E1000_DEV_ID_I354_BACKPLANE_2_5GBPS, "Intel(R) I354 (2.5 GbE Backplane)"),
243         PVID(0x8086, E1000_DEV_ID_I354_SGMII, "Intel(R) I354 (SGMII)"),
244         /* required last entry */
245         PVID_END
246 };
247
248 /*********************************************************************
249  *  Function prototypes
250  *********************************************************************/
251 static void     *em_register(device_t);
252 static void     *igb_register(device_t);
253 static int      em_if_attach_pre(if_ctx_t);
254 static int      em_if_attach_post(if_ctx_t);
255 static int      em_if_detach(if_ctx_t);
256 static int      em_if_shutdown(if_ctx_t);
257 static int      em_if_suspend(if_ctx_t);
258 static int      em_if_resume(if_ctx_t);
259
260 static int      em_if_tx_queues_alloc(if_ctx_t, caddr_t *, uint64_t *, int, int);
261 static int      em_if_rx_queues_alloc(if_ctx_t, caddr_t *, uint64_t *, int, int);
262 static void     em_if_queues_free(if_ctx_t);
263
264 static uint64_t em_if_get_counter(if_ctx_t, ift_counter);
265 static void     em_if_init(if_ctx_t);
266 static void     em_if_stop(if_ctx_t);
267 static void     em_if_media_status(if_ctx_t, struct ifmediareq *);
268 static int      em_if_media_change(if_ctx_t);
269 static int      em_if_mtu_set(if_ctx_t, uint32_t);
270 static void     em_if_timer(if_ctx_t, uint16_t);
271 static void     em_if_vlan_register(if_ctx_t, u16);
272 static void     em_if_vlan_unregister(if_ctx_t, u16);
273 static void     em_if_watchdog_reset(if_ctx_t);
274 static bool     em_if_needs_restart(if_ctx_t, enum iflib_restart_event);
275
276 static void     em_identify_hardware(if_ctx_t);
277 static int      em_allocate_pci_resources(if_ctx_t);
278 static void     em_free_pci_resources(if_ctx_t);
279 static void     em_reset(if_ctx_t);
280 static int      em_setup_interface(if_ctx_t);
281 static int      em_setup_msix(if_ctx_t);
282
283 static void     em_initialize_transmit_unit(if_ctx_t);
284 static void     em_initialize_receive_unit(if_ctx_t);
285
286 static void     em_if_intr_enable(if_ctx_t);
287 static void     em_if_intr_disable(if_ctx_t);
288 static void     igb_if_intr_enable(if_ctx_t);
289 static void     igb_if_intr_disable(if_ctx_t);
290 static int      em_if_rx_queue_intr_enable(if_ctx_t, uint16_t);
291 static int      em_if_tx_queue_intr_enable(if_ctx_t, uint16_t);
292 static int      igb_if_rx_queue_intr_enable(if_ctx_t, uint16_t);
293 static int      igb_if_tx_queue_intr_enable(if_ctx_t, uint16_t);
294 static void     em_if_multi_set(if_ctx_t);
295 static void     em_if_update_admin_status(if_ctx_t);
296 static void     em_if_debug(if_ctx_t);
297 static void     em_update_stats_counters(struct e1000_softc *);
298 static void     em_add_hw_stats(struct e1000_softc *);
299 static int      em_if_set_promisc(if_ctx_t, int);
300 static bool     em_if_vlan_filter_capable(if_ctx_t);
301 static bool     em_if_vlan_filter_used(if_ctx_t);
302 static void     em_if_vlan_filter_enable(struct e1000_softc *);
303 static void     em_if_vlan_filter_disable(struct e1000_softc *);
304 static void     em_if_vlan_filter_write(struct e1000_softc *);
305 static void     em_setup_vlan_hw_support(if_ctx_t ctx);
306 static int      em_sysctl_nvm_info(SYSCTL_HANDLER_ARGS);
307 static void     em_print_nvm_info(struct e1000_softc *);
308 static void     em_fw_version_locked(if_ctx_t);
309 static void     em_sbuf_fw_version(struct e1000_fw_version *, struct sbuf *);
310 static void     em_print_fw_version(struct e1000_softc *);
311 static int      em_sysctl_print_fw_version(SYSCTL_HANDLER_ARGS);
312 static int      em_sysctl_debug_info(SYSCTL_HANDLER_ARGS);
313 static int      em_get_rs(SYSCTL_HANDLER_ARGS);
314 static void     em_print_debug_info(struct e1000_softc *);
315 static int      em_is_valid_ether_addr(u8 *);
316 static int      em_sysctl_int_delay(SYSCTL_HANDLER_ARGS);
317 static void     em_add_int_delay_sysctl(struct e1000_softc *, const char *,
318                     const char *, struct em_int_delay_info *, int, int);
319 /* Management and WOL Support */
320 static void     em_init_manageability(struct e1000_softc *);
321 static void     em_release_manageability(struct e1000_softc *);
322 static void     em_get_hw_control(struct e1000_softc *);
323 static void     em_release_hw_control(struct e1000_softc *);
324 static void     em_get_wakeup(if_ctx_t);
325 static void     em_enable_wakeup(if_ctx_t);
326 static int      em_enable_phy_wakeup(struct e1000_softc *);
327 static void     em_disable_aspm(struct e1000_softc *);
328
329 int             em_intr(void *);
330
331 /* MSI-X handlers */
332 static int      em_if_msix_intr_assign(if_ctx_t, int);
333 static int      em_msix_link(void *);
334 static void     em_handle_link(void *);
335
336 static void     em_enable_vectors_82574(if_ctx_t);
337
338 static int      em_set_flowcntl(SYSCTL_HANDLER_ARGS);
339 static int      em_sysctl_eee(SYSCTL_HANDLER_ARGS);
340 static void     em_if_led_func(if_ctx_t, int);
341
342 static int      em_get_regs(SYSCTL_HANDLER_ARGS);
343
344 static void     lem_smartspeed(struct e1000_softc *);
345 static void     igb_configure_queues(struct e1000_softc *);
346
347
348 /*********************************************************************
349  *  FreeBSD Device Interface Entry Points
350  *********************************************************************/
351 static device_method_t em_methods[] = {
352         /* Device interface */
353         DEVMETHOD(device_register, em_register),
354         DEVMETHOD(device_probe, iflib_device_probe),
355         DEVMETHOD(device_attach, iflib_device_attach),
356         DEVMETHOD(device_detach, iflib_device_detach),
357         DEVMETHOD(device_shutdown, iflib_device_shutdown),
358         DEVMETHOD(device_suspend, iflib_device_suspend),
359         DEVMETHOD(device_resume, iflib_device_resume),
360         DEVMETHOD_END
361 };
362
363 static device_method_t igb_methods[] = {
364         /* Device interface */
365         DEVMETHOD(device_register, igb_register),
366         DEVMETHOD(device_probe, iflib_device_probe),
367         DEVMETHOD(device_attach, iflib_device_attach),
368         DEVMETHOD(device_detach, iflib_device_detach),
369         DEVMETHOD(device_shutdown, iflib_device_shutdown),
370         DEVMETHOD(device_suspend, iflib_device_suspend),
371         DEVMETHOD(device_resume, iflib_device_resume),
372         DEVMETHOD_END
373 };
374
375
376 static driver_t em_driver = {
377         "em", em_methods, sizeof(struct e1000_softc),
378 };
379
380 DRIVER_MODULE(em, pci, em_driver, 0, 0);
381
382 MODULE_DEPEND(em, pci, 1, 1, 1);
383 MODULE_DEPEND(em, ether, 1, 1, 1);
384 MODULE_DEPEND(em, iflib, 1, 1, 1);
385
386 IFLIB_PNP_INFO(pci, em, em_vendor_info_array);
387
388 static driver_t igb_driver = {
389         "igb", igb_methods, sizeof(struct e1000_softc),
390 };
391
392 DRIVER_MODULE(igb, pci, igb_driver, 0, 0);
393
394 MODULE_DEPEND(igb, pci, 1, 1, 1);
395 MODULE_DEPEND(igb, ether, 1, 1, 1);
396 MODULE_DEPEND(igb, iflib, 1, 1, 1);
397
398 IFLIB_PNP_INFO(pci, igb, igb_vendor_info_array);
399
400 static device_method_t em_if_methods[] = {
401         DEVMETHOD(ifdi_attach_pre, em_if_attach_pre),
402         DEVMETHOD(ifdi_attach_post, em_if_attach_post),
403         DEVMETHOD(ifdi_detach, em_if_detach),
404         DEVMETHOD(ifdi_shutdown, em_if_shutdown),
405         DEVMETHOD(ifdi_suspend, em_if_suspend),
406         DEVMETHOD(ifdi_resume, em_if_resume),
407         DEVMETHOD(ifdi_init, em_if_init),
408         DEVMETHOD(ifdi_stop, em_if_stop),
409         DEVMETHOD(ifdi_msix_intr_assign, em_if_msix_intr_assign),
410         DEVMETHOD(ifdi_intr_enable, em_if_intr_enable),
411         DEVMETHOD(ifdi_intr_disable, em_if_intr_disable),
412         DEVMETHOD(ifdi_tx_queues_alloc, em_if_tx_queues_alloc),
413         DEVMETHOD(ifdi_rx_queues_alloc, em_if_rx_queues_alloc),
414         DEVMETHOD(ifdi_queues_free, em_if_queues_free),
415         DEVMETHOD(ifdi_update_admin_status, em_if_update_admin_status),
416         DEVMETHOD(ifdi_multi_set, em_if_multi_set),
417         DEVMETHOD(ifdi_media_status, em_if_media_status),
418         DEVMETHOD(ifdi_media_change, em_if_media_change),
419         DEVMETHOD(ifdi_mtu_set, em_if_mtu_set),
420         DEVMETHOD(ifdi_promisc_set, em_if_set_promisc),
421         DEVMETHOD(ifdi_timer, em_if_timer),
422         DEVMETHOD(ifdi_watchdog_reset, em_if_watchdog_reset),
423         DEVMETHOD(ifdi_vlan_register, em_if_vlan_register),
424         DEVMETHOD(ifdi_vlan_unregister, em_if_vlan_unregister),
425         DEVMETHOD(ifdi_get_counter, em_if_get_counter),
426         DEVMETHOD(ifdi_led_func, em_if_led_func),
427         DEVMETHOD(ifdi_rx_queue_intr_enable, em_if_rx_queue_intr_enable),
428         DEVMETHOD(ifdi_tx_queue_intr_enable, em_if_tx_queue_intr_enable),
429         DEVMETHOD(ifdi_debug, em_if_debug),
430         DEVMETHOD(ifdi_needs_restart, em_if_needs_restart),
431         DEVMETHOD_END
432 };
433
434 static driver_t em_if_driver = {
435         "em_if", em_if_methods, sizeof(struct e1000_softc)
436 };
437
438 static device_method_t igb_if_methods[] = {
439         DEVMETHOD(ifdi_attach_pre, em_if_attach_pre),
440         DEVMETHOD(ifdi_attach_post, em_if_attach_post),
441         DEVMETHOD(ifdi_detach, em_if_detach),
442         DEVMETHOD(ifdi_shutdown, em_if_shutdown),
443         DEVMETHOD(ifdi_suspend, em_if_suspend),
444         DEVMETHOD(ifdi_resume, em_if_resume),
445         DEVMETHOD(ifdi_init, em_if_init),
446         DEVMETHOD(ifdi_stop, em_if_stop),
447         DEVMETHOD(ifdi_msix_intr_assign, em_if_msix_intr_assign),
448         DEVMETHOD(ifdi_intr_enable, igb_if_intr_enable),
449         DEVMETHOD(ifdi_intr_disable, igb_if_intr_disable),
450         DEVMETHOD(ifdi_tx_queues_alloc, em_if_tx_queues_alloc),
451         DEVMETHOD(ifdi_rx_queues_alloc, em_if_rx_queues_alloc),
452         DEVMETHOD(ifdi_queues_free, em_if_queues_free),
453         DEVMETHOD(ifdi_update_admin_status, em_if_update_admin_status),
454         DEVMETHOD(ifdi_multi_set, em_if_multi_set),
455         DEVMETHOD(ifdi_media_status, em_if_media_status),
456         DEVMETHOD(ifdi_media_change, em_if_media_change),
457         DEVMETHOD(ifdi_mtu_set, em_if_mtu_set),
458         DEVMETHOD(ifdi_promisc_set, em_if_set_promisc),
459         DEVMETHOD(ifdi_timer, em_if_timer),
460         DEVMETHOD(ifdi_watchdog_reset, em_if_watchdog_reset),
461         DEVMETHOD(ifdi_vlan_register, em_if_vlan_register),
462         DEVMETHOD(ifdi_vlan_unregister, em_if_vlan_unregister),
463         DEVMETHOD(ifdi_get_counter, em_if_get_counter),
464         DEVMETHOD(ifdi_led_func, em_if_led_func),
465         DEVMETHOD(ifdi_rx_queue_intr_enable, igb_if_rx_queue_intr_enable),
466         DEVMETHOD(ifdi_tx_queue_intr_enable, igb_if_tx_queue_intr_enable),
467         DEVMETHOD(ifdi_debug, em_if_debug),
468         DEVMETHOD(ifdi_needs_restart, em_if_needs_restart),
469         DEVMETHOD_END
470 };
471
472 static driver_t igb_if_driver = {
473         "igb_if", igb_if_methods, sizeof(struct e1000_softc)
474 };
475
476 /*********************************************************************
477  *  Tunable default values.
478  *********************************************************************/
479
480 #define EM_TICKS_TO_USECS(ticks)        ((1024 * (ticks) + 500) / 1000)
481 #define EM_USECS_TO_TICKS(usecs)        ((1000 * (usecs) + 512) / 1024)
482
483 #define MAX_INTS_PER_SEC        8000
484 #define DEFAULT_ITR             (1000000000/(MAX_INTS_PER_SEC * 256))
485
486 /* Allow common code without TSO */
487 #ifndef CSUM_TSO
488 #define CSUM_TSO        0
489 #endif
490
491 static SYSCTL_NODE(_hw, OID_AUTO, em, CTLFLAG_RD | CTLFLAG_MPSAFE, 0,
492     "EM driver parameters");
493
494 static int em_disable_crc_stripping = 0;
495 SYSCTL_INT(_hw_em, OID_AUTO, disable_crc_stripping, CTLFLAG_RDTUN,
496     &em_disable_crc_stripping, 0, "Disable CRC Stripping");
497
498 static int em_tx_int_delay_dflt = EM_TICKS_TO_USECS(EM_TIDV);
499 static int em_rx_int_delay_dflt = EM_TICKS_TO_USECS(EM_RDTR);
500 SYSCTL_INT(_hw_em, OID_AUTO, tx_int_delay, CTLFLAG_RDTUN, &em_tx_int_delay_dflt,
501     0, "Default transmit interrupt delay in usecs");
502 SYSCTL_INT(_hw_em, OID_AUTO, rx_int_delay, CTLFLAG_RDTUN, &em_rx_int_delay_dflt,
503     0, "Default receive interrupt delay in usecs");
504
505 static int em_tx_abs_int_delay_dflt = EM_TICKS_TO_USECS(EM_TADV);
506 static int em_rx_abs_int_delay_dflt = EM_TICKS_TO_USECS(EM_RADV);
507 SYSCTL_INT(_hw_em, OID_AUTO, tx_abs_int_delay, CTLFLAG_RDTUN,
508     &em_tx_abs_int_delay_dflt, 0,
509     "Default transmit interrupt delay limit in usecs");
510 SYSCTL_INT(_hw_em, OID_AUTO, rx_abs_int_delay, CTLFLAG_RDTUN,
511     &em_rx_abs_int_delay_dflt, 0,
512     "Default receive interrupt delay limit in usecs");
513
514 static int em_smart_pwr_down = false;
515 SYSCTL_INT(_hw_em, OID_AUTO, smart_pwr_down, CTLFLAG_RDTUN, &em_smart_pwr_down,
516     0, "Set to true to leave smart power down enabled on newer adapters");
517
518 /* Controls whether promiscuous also shows bad packets */
519 static int em_debug_sbp = false;
520 SYSCTL_INT(_hw_em, OID_AUTO, sbp, CTLFLAG_RDTUN, &em_debug_sbp, 0,
521     "Show bad packets in promiscuous mode");
522
523 /* How many packets rxeof tries to clean at a time */
524 static int em_rx_process_limit = 100;
525 SYSCTL_INT(_hw_em, OID_AUTO, rx_process_limit, CTLFLAG_RDTUN,
526     &em_rx_process_limit, 0,
527     "Maximum number of received packets to process "
528     "at a time, -1 means unlimited");
529
530 /* Energy efficient ethernet - default to OFF */
531 static int eee_setting = 1;
532 SYSCTL_INT(_hw_em, OID_AUTO, eee_setting, CTLFLAG_RDTUN, &eee_setting, 0,
533     "Enable Energy Efficient Ethernet");
534
535 /*
536 ** Tuneable Interrupt rate
537 */
538 static int em_max_interrupt_rate = 8000;
539 SYSCTL_INT(_hw_em, OID_AUTO, max_interrupt_rate, CTLFLAG_RDTUN,
540     &em_max_interrupt_rate, 0, "Maximum interrupts per second");
541
542
543
544 /* Global used in WOL setup with multiport cards */
545 static int global_quad_port_a = 0;
546
547 extern struct if_txrx igb_txrx;
548 extern struct if_txrx em_txrx;
549 extern struct if_txrx lem_txrx;
550
551 static struct if_shared_ctx em_sctx_init = {
552         .isc_magic = IFLIB_MAGIC,
553         .isc_q_align = PAGE_SIZE,
554         .isc_tx_maxsize = EM_TSO_SIZE + sizeof(struct ether_vlan_header),
555         .isc_tx_maxsegsize = PAGE_SIZE,
556         .isc_tso_maxsize = EM_TSO_SIZE + sizeof(struct ether_vlan_header),
557         .isc_tso_maxsegsize = EM_TSO_SEG_SIZE,
558         .isc_rx_maxsize = MJUM9BYTES,
559         .isc_rx_nsegments = 1,
560         .isc_rx_maxsegsize = MJUM9BYTES,
561         .isc_nfl = 1,
562         .isc_nrxqs = 1,
563         .isc_ntxqs = 1,
564         .isc_admin_intrcnt = 1,
565         .isc_vendor_info = em_vendor_info_array,
566         .isc_driver_version = em_driver_version,
567         .isc_driver = &em_if_driver,
568         .isc_flags = IFLIB_NEED_SCRATCH | IFLIB_TSO_INIT_IP | IFLIB_NEED_ZERO_CSUM,
569
570         .isc_nrxd_min = {EM_MIN_RXD},
571         .isc_ntxd_min = {EM_MIN_TXD},
572         .isc_nrxd_max = {EM_MAX_RXD},
573         .isc_ntxd_max = {EM_MAX_TXD},
574         .isc_nrxd_default = {EM_DEFAULT_RXD},
575         .isc_ntxd_default = {EM_DEFAULT_TXD},
576 };
577
578 static struct if_shared_ctx igb_sctx_init = {
579         .isc_magic = IFLIB_MAGIC,
580         .isc_q_align = PAGE_SIZE,
581         .isc_tx_maxsize = EM_TSO_SIZE + sizeof(struct ether_vlan_header),
582         .isc_tx_maxsegsize = PAGE_SIZE,
583         .isc_tso_maxsize = EM_TSO_SIZE + sizeof(struct ether_vlan_header),
584         .isc_tso_maxsegsize = EM_TSO_SEG_SIZE,
585         .isc_rx_maxsize = MJUM9BYTES,
586         .isc_rx_nsegments = 1,
587         .isc_rx_maxsegsize = MJUM9BYTES,
588         .isc_nfl = 1,
589         .isc_nrxqs = 1,
590         .isc_ntxqs = 1,
591         .isc_admin_intrcnt = 1,
592         .isc_vendor_info = igb_vendor_info_array,
593         .isc_driver_version = em_driver_version,
594         .isc_driver = &igb_if_driver,
595         .isc_flags = IFLIB_NEED_SCRATCH | IFLIB_TSO_INIT_IP | IFLIB_NEED_ZERO_CSUM,
596
597         .isc_nrxd_min = {EM_MIN_RXD},
598         .isc_ntxd_min = {EM_MIN_TXD},
599         .isc_nrxd_max = {IGB_MAX_RXD},
600         .isc_ntxd_max = {IGB_MAX_TXD},
601         .isc_nrxd_default = {EM_DEFAULT_RXD},
602         .isc_ntxd_default = {EM_DEFAULT_TXD},
603 };
604
605 /*****************************************************************
606  *
607  * Dump Registers
608  *
609  ****************************************************************/
610 #define IGB_REGS_LEN 739
611
612 static int em_get_regs(SYSCTL_HANDLER_ARGS)
613 {
614         struct e1000_softc *sc = (struct e1000_softc *)arg1;
615         struct e1000_hw *hw = &sc->hw;
616         struct sbuf *sb;
617         u32 *regs_buff;
618         int rc;
619
620         regs_buff = malloc(sizeof(u32) * IGB_REGS_LEN, M_DEVBUF, M_WAITOK);
621         memset(regs_buff, 0, IGB_REGS_LEN * sizeof(u32));
622
623         rc = sysctl_wire_old_buffer(req, 0);
624         MPASS(rc == 0);
625         if (rc != 0) {
626                 free(regs_buff, M_DEVBUF);
627                 return (rc);
628         }
629
630         sb = sbuf_new_for_sysctl(NULL, NULL, 32*400, req);
631         MPASS(sb != NULL);
632         if (sb == NULL) {
633                 free(regs_buff, M_DEVBUF);
634                 return (ENOMEM);
635         }
636
637         /* General Registers */
638         regs_buff[0] = E1000_READ_REG(hw, E1000_CTRL);
639         regs_buff[1] = E1000_READ_REG(hw, E1000_STATUS);
640         regs_buff[2] = E1000_READ_REG(hw, E1000_CTRL_EXT);
641         regs_buff[3] = E1000_READ_REG(hw, E1000_ICR);
642         regs_buff[4] = E1000_READ_REG(hw, E1000_RCTL);
643         regs_buff[5] = E1000_READ_REG(hw, E1000_RDLEN(0));
644         regs_buff[6] = E1000_READ_REG(hw, E1000_RDH(0));
645         regs_buff[7] = E1000_READ_REG(hw, E1000_RDT(0));
646         regs_buff[8] = E1000_READ_REG(hw, E1000_RXDCTL(0));
647         regs_buff[9] = E1000_READ_REG(hw, E1000_RDBAL(0));
648         regs_buff[10] = E1000_READ_REG(hw, E1000_RDBAH(0));
649         regs_buff[11] = E1000_READ_REG(hw, E1000_TCTL);
650         regs_buff[12] = E1000_READ_REG(hw, E1000_TDBAL(0));
651         regs_buff[13] = E1000_READ_REG(hw, E1000_TDBAH(0));
652         regs_buff[14] = E1000_READ_REG(hw, E1000_TDLEN(0));
653         regs_buff[15] = E1000_READ_REG(hw, E1000_TDH(0));
654         regs_buff[16] = E1000_READ_REG(hw, E1000_TDT(0));
655         regs_buff[17] = E1000_READ_REG(hw, E1000_TXDCTL(0));
656         regs_buff[18] = E1000_READ_REG(hw, E1000_TDFH);
657         regs_buff[19] = E1000_READ_REG(hw, E1000_TDFT);
658         regs_buff[20] = E1000_READ_REG(hw, E1000_TDFHS);
659         regs_buff[21] = E1000_READ_REG(hw, E1000_TDFPC);
660
661         sbuf_printf(sb, "General Registers\n");
662         sbuf_printf(sb, "\tCTRL\t %08x\n", regs_buff[0]);
663         sbuf_printf(sb, "\tSTATUS\t %08x\n", regs_buff[1]);
664         sbuf_printf(sb, "\tCTRL_EXT\t %08x\n\n", regs_buff[2]);
665
666         sbuf_printf(sb, "Interrupt Registers\n");
667         sbuf_printf(sb, "\tICR\t %08x\n\n", regs_buff[3]);
668
669         sbuf_printf(sb, "RX Registers\n");
670         sbuf_printf(sb, "\tRCTL\t %08x\n", regs_buff[4]);
671         sbuf_printf(sb, "\tRDLEN\t %08x\n", regs_buff[5]);
672         sbuf_printf(sb, "\tRDH\t %08x\n", regs_buff[6]);
673         sbuf_printf(sb, "\tRDT\t %08x\n", regs_buff[7]);
674         sbuf_printf(sb, "\tRXDCTL\t %08x\n", regs_buff[8]);
675         sbuf_printf(sb, "\tRDBAL\t %08x\n", regs_buff[9]);
676         sbuf_printf(sb, "\tRDBAH\t %08x\n\n", regs_buff[10]);
677
678         sbuf_printf(sb, "TX Registers\n");
679         sbuf_printf(sb, "\tTCTL\t %08x\n", regs_buff[11]);
680         sbuf_printf(sb, "\tTDBAL\t %08x\n", regs_buff[12]);
681         sbuf_printf(sb, "\tTDBAH\t %08x\n", regs_buff[13]);
682         sbuf_printf(sb, "\tTDLEN\t %08x\n", regs_buff[14]);
683         sbuf_printf(sb, "\tTDH\t %08x\n", regs_buff[15]);
684         sbuf_printf(sb, "\tTDT\t %08x\n", regs_buff[16]);
685         sbuf_printf(sb, "\tTXDCTL\t %08x\n", regs_buff[17]);
686         sbuf_printf(sb, "\tTDFH\t %08x\n", regs_buff[18]);
687         sbuf_printf(sb, "\tTDFT\t %08x\n", regs_buff[19]);
688         sbuf_printf(sb, "\tTDFHS\t %08x\n", regs_buff[20]);
689         sbuf_printf(sb, "\tTDFPC\t %08x\n\n", regs_buff[21]);
690
691         free(regs_buff, M_DEVBUF);
692
693 #ifdef DUMP_DESCS
694         {
695                 if_softc_ctx_t scctx = sc->shared;
696                 struct rx_ring *rxr = &rx_que->rxr;
697                 struct tx_ring *txr = &tx_que->txr;
698                 int ntxd = scctx->isc_ntxd[0];
699                 int nrxd = scctx->isc_nrxd[0];
700                 int j;
701
702         for (j = 0; j < nrxd; j++) {
703                 u32 staterr = le32toh(rxr->rx_base[j].wb.upper.status_error);
704                 u32 length =  le32toh(rxr->rx_base[j].wb.upper.length);
705                 sbuf_printf(sb, "\tReceive Descriptor Address %d: %08" PRIx64 "  Error:%d  Length:%d\n", j, rxr->rx_base[j].read.buffer_addr, staterr, length);
706         }
707
708         for (j = 0; j < min(ntxd, 256); j++) {
709                 unsigned int *ptr = (unsigned int *)&txr->tx_base[j];
710
711                 sbuf_printf(sb, "\tTXD[%03d] [0]: %08x [1]: %08x [2]: %08x [3]: %08x  eop: %d DD=%d\n",
712                             j, ptr[0], ptr[1], ptr[2], ptr[3], buf->eop,
713                             buf->eop != -1 ? txr->tx_base[buf->eop].upper.fields.status & E1000_TXD_STAT_DD : 0);
714
715         }
716         }
717 #endif
718
719         rc = sbuf_finish(sb);
720         sbuf_delete(sb);
721         return(rc);
722 }
723
724 static void *
725 em_register(device_t dev)
726 {
727         return (&em_sctx_init);
728 }
729
730 static void *
731 igb_register(device_t dev)
732 {
733         return (&igb_sctx_init);
734 }
735
736 static int
737 em_set_num_queues(if_ctx_t ctx)
738 {
739         struct e1000_softc *sc = iflib_get_softc(ctx);
740         int maxqueues;
741
742         /* Sanity check based on HW */
743         switch (sc->hw.mac.type) {
744         case e1000_82576:
745         case e1000_82580:
746         case e1000_i350:
747         case e1000_i354:
748                 maxqueues = 8;
749                 break;
750         case e1000_i210:
751         case e1000_82575:
752                 maxqueues = 4;
753                 break;
754         case e1000_i211:
755         case e1000_82574:
756                 maxqueues = 2;
757                 break;
758         default:
759                 maxqueues = 1;
760                 break;
761         }
762
763         return (maxqueues);
764 }
765
766 #define LEM_CAPS                                                        \
767     IFCAP_HWCSUM | IFCAP_VLAN_MTU | IFCAP_VLAN_HWTAGGING |              \
768     IFCAP_VLAN_HWCSUM | IFCAP_WOL | IFCAP_VLAN_HWFILTER
769
770 #define EM_CAPS                                                         \
771     IFCAP_HWCSUM | IFCAP_VLAN_MTU | IFCAP_VLAN_HWTAGGING |              \
772     IFCAP_VLAN_HWCSUM | IFCAP_WOL | IFCAP_VLAN_HWFILTER | IFCAP_TSO4 |  \
773     IFCAP_LRO | IFCAP_VLAN_HWTSO
774
775 #define IGB_CAPS                                                        \
776     IFCAP_HWCSUM | IFCAP_VLAN_MTU | IFCAP_VLAN_HWTAGGING |              \
777     IFCAP_VLAN_HWCSUM | IFCAP_WOL | IFCAP_VLAN_HWFILTER | IFCAP_TSO4 |  \
778     IFCAP_LRO | IFCAP_VLAN_HWTSO | IFCAP_JUMBO_MTU | IFCAP_HWCSUM_IPV6 |\
779     IFCAP_TSO6
780
781 /*********************************************************************
782  *  Device initialization routine
783  *
784  *  The attach entry point is called when the driver is being loaded.
785  *  This routine identifies the type of hardware, allocates all resources
786  *  and initializes the hardware.
787  *
788  *  return 0 on success, positive on failure
789  *********************************************************************/
790 static int
791 em_if_attach_pre(if_ctx_t ctx)
792 {
793         struct e1000_softc *sc;
794         if_softc_ctx_t scctx;
795         device_t dev;
796         struct e1000_hw *hw;
797         struct sysctl_oid_list *child;
798         struct sysctl_ctx_list *ctx_list;
799         int error = 0;
800
801         INIT_DEBUGOUT("em_if_attach_pre: begin");
802         dev = iflib_get_dev(ctx);
803         sc = iflib_get_softc(ctx);
804
805         sc->ctx = sc->osdep.ctx = ctx;
806         sc->dev = sc->osdep.dev = dev;
807         scctx = sc->shared = iflib_get_softc_ctx(ctx);
808         sc->media = iflib_get_media(ctx);
809         hw = &sc->hw;
810
811         sc->tx_process_limit = scctx->isc_ntxd[0];
812
813         /* Determine hardware and mac info */
814         em_identify_hardware(ctx);
815
816         /* SYSCTL stuff */
817         ctx_list = device_get_sysctl_ctx(dev);
818         child = SYSCTL_CHILDREN(device_get_sysctl_tree(dev));
819
820         SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "nvm",
821             CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT, sc, 0,
822             em_sysctl_nvm_info, "I", "NVM Information");
823
824         SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "fw_version",
825             CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_NEEDGIANT, sc, 0,
826             em_sysctl_print_fw_version, "A",
827             "Prints FW/NVM Versions");
828
829         SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "debug",
830             CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT, sc, 0,
831             em_sysctl_debug_info, "I", "Debug Information");
832
833         SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "fc",
834             CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT, sc, 0,
835             em_set_flowcntl, "I", "Flow Control");
836
837         SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "reg_dump",
838             CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_NEEDGIANT, sc, 0,
839             em_get_regs, "A", "Dump Registers");
840
841         SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "rs_dump",
842             CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT, sc, 0,
843             em_get_rs, "I", "Dump RS indexes");
844
845         scctx->isc_tx_nsegments = EM_MAX_SCATTER;
846         scctx->isc_nrxqsets_max = scctx->isc_ntxqsets_max = em_set_num_queues(ctx);
847         if (bootverbose)
848                 device_printf(dev, "attach_pre capping queues at %d\n",
849                     scctx->isc_ntxqsets_max);
850
851         if (hw->mac.type >= igb_mac_min) {
852                 scctx->isc_txqsizes[0] = roundup2(scctx->isc_ntxd[0] * sizeof(union e1000_adv_tx_desc), EM_DBA_ALIGN);
853                 scctx->isc_rxqsizes[0] = roundup2(scctx->isc_nrxd[0] * sizeof(union e1000_adv_rx_desc), EM_DBA_ALIGN);
854                 scctx->isc_txd_size[0] = sizeof(union e1000_adv_tx_desc);
855                 scctx->isc_rxd_size[0] = sizeof(union e1000_adv_rx_desc);
856                 scctx->isc_txrx = &igb_txrx;
857                 scctx->isc_tx_tso_segments_max = EM_MAX_SCATTER;
858                 scctx->isc_tx_tso_size_max = EM_TSO_SIZE;
859                 scctx->isc_tx_tso_segsize_max = EM_TSO_SEG_SIZE;
860                 scctx->isc_capabilities = scctx->isc_capenable = IGB_CAPS;
861                 scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP | CSUM_TSO |
862                      CSUM_IP6_TCP | CSUM_IP6_UDP;
863                 if (hw->mac.type != e1000_82575)
864                         scctx->isc_tx_csum_flags |= CSUM_SCTP | CSUM_IP6_SCTP;
865                 /*
866                 ** Some new devices, as with ixgbe, now may
867                 ** use a different BAR, so we need to keep
868                 ** track of which is used.
869                 */
870                 scctx->isc_msix_bar = pci_msix_table_bar(dev);
871         } else if (hw->mac.type >= em_mac_min) {
872                 scctx->isc_txqsizes[0] = roundup2(scctx->isc_ntxd[0]* sizeof(struct e1000_tx_desc), EM_DBA_ALIGN);
873                 scctx->isc_rxqsizes[0] = roundup2(scctx->isc_nrxd[0] * sizeof(union e1000_rx_desc_extended), EM_DBA_ALIGN);
874                 scctx->isc_txd_size[0] = sizeof(struct e1000_tx_desc);
875                 scctx->isc_rxd_size[0] = sizeof(union e1000_rx_desc_extended);
876                 scctx->isc_txrx = &em_txrx;
877                 scctx->isc_tx_tso_segments_max = EM_MAX_SCATTER;
878                 scctx->isc_tx_tso_size_max = EM_TSO_SIZE;
879                 scctx->isc_tx_tso_segsize_max = EM_TSO_SEG_SIZE;
880                 scctx->isc_capabilities = scctx->isc_capenable = EM_CAPS;
881                 /*
882                  * For EM-class devices, don't enable IFCAP_{TSO4,VLAN_HWTSO}
883                  * by default as we don't have workarounds for all associated
884                  * silicon errata.  E. g., with several MACs such as 82573E,
885                  * TSO only works at Gigabit speed and otherwise can cause the
886                  * hardware to hang (which also would be next to impossible to
887                  * work around given that already queued TSO-using descriptors
888                  * would need to be flushed and vlan(4) reconfigured at runtime
889                  * in case of a link speed change).  Moreover, MACs like 82579
890                  * still can hang at Gigabit even with all publicly documented
891                  * TSO workarounds implemented.  Generally, the penality of
892                  * these workarounds is rather high and may involve copying
893                  * mbuf data around so advantages of TSO lapse.  Still, TSO may
894                  * work for a few MACs of this class - at least when sticking
895                  * with Gigabit - in which case users may enable TSO manually.
896                  */
897                 scctx->isc_capenable &= ~(IFCAP_TSO4 | IFCAP_VLAN_HWTSO);
898                 scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP | CSUM_IP_TSO;
899                 /*
900                  * We support MSI-X with 82574 only, but indicate to iflib(4)
901                  * that it shall give MSI at least a try with other devices.
902                  */
903                 if (hw->mac.type == e1000_82574) {
904                         scctx->isc_msix_bar = pci_msix_table_bar(dev);;
905                 } else {
906                         scctx->isc_msix_bar = -1;
907                         scctx->isc_disable_msix = 1;
908                 }
909         } else {
910                 scctx->isc_txqsizes[0] = roundup2((scctx->isc_ntxd[0] + 1) * sizeof(struct e1000_tx_desc), EM_DBA_ALIGN);
911                 scctx->isc_rxqsizes[0] = roundup2((scctx->isc_nrxd[0] + 1) * sizeof(struct e1000_rx_desc), EM_DBA_ALIGN);
912                 scctx->isc_txd_size[0] = sizeof(struct e1000_tx_desc);
913                 scctx->isc_rxd_size[0] = sizeof(struct e1000_rx_desc);
914                 scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP;
915                 scctx->isc_txrx = &lem_txrx;
916                 scctx->isc_capabilities = LEM_CAPS;
917                 if (hw->mac.type < e1000_82543)
918                         scctx->isc_capabilities &= ~(IFCAP_HWCSUM|IFCAP_VLAN_HWCSUM);
919                 /* 82541ER doesn't do HW tagging */
920                 if (hw->device_id == E1000_DEV_ID_82541ER || hw->device_id == E1000_DEV_ID_82541ER_LOM)
921                         scctx->isc_capabilities &= ~IFCAP_VLAN_HWTAGGING;
922                 /* INTx only */
923                 scctx->isc_msix_bar = 0;
924                 scctx->isc_capenable = scctx->isc_capabilities;
925         }
926
927         /* Setup PCI resources */
928         if (em_allocate_pci_resources(ctx)) {
929                 device_printf(dev, "Allocation of PCI resources failed\n");
930                 error = ENXIO;
931                 goto err_pci;
932         }
933
934         /*
935         ** For ICH8 and family we need to
936         ** map the flash memory, and this
937         ** must happen after the MAC is
938         ** identified
939         */
940         if ((hw->mac.type == e1000_ich8lan) ||
941             (hw->mac.type == e1000_ich9lan) ||
942             (hw->mac.type == e1000_ich10lan) ||
943             (hw->mac.type == e1000_pchlan) ||
944             (hw->mac.type == e1000_pch2lan) ||
945             (hw->mac.type == e1000_pch_lpt)) {
946                 int rid = EM_BAR_TYPE_FLASH;
947                 sc->flash = bus_alloc_resource_any(dev,
948                     SYS_RES_MEMORY, &rid, RF_ACTIVE);
949                 if (sc->flash == NULL) {
950                         device_printf(dev, "Mapping of Flash failed\n");
951                         error = ENXIO;
952                         goto err_pci;
953                 }
954                 /* This is used in the shared code */
955                 hw->flash_address = (u8 *)sc->flash;
956                 sc->osdep.flash_bus_space_tag =
957                     rman_get_bustag(sc->flash);
958                 sc->osdep.flash_bus_space_handle =
959                     rman_get_bushandle(sc->flash);
960         }
961         /*
962         ** In the new SPT device flash is not  a
963         ** separate BAR, rather it is also in BAR0,
964         ** so use the same tag and an offset handle for the
965         ** FLASH read/write macros in the shared code.
966         */
967         else if (hw->mac.type >= e1000_pch_spt) {
968                 sc->osdep.flash_bus_space_tag =
969                     sc->osdep.mem_bus_space_tag;
970                 sc->osdep.flash_bus_space_handle =
971                     sc->osdep.mem_bus_space_handle
972                     + E1000_FLASH_BASE_ADDR;
973         }
974
975         /* Do Shared Code initialization */
976         error = e1000_setup_init_funcs(hw, true);
977         if (error) {
978                 device_printf(dev, "Setup of Shared code failed, error %d\n",
979                     error);
980                 error = ENXIO;
981                 goto err_pci;
982         }
983
984         em_setup_msix(ctx);
985         e1000_get_bus_info(hw);
986
987         /* Set up some sysctls for the tunable interrupt delays */
988         em_add_int_delay_sysctl(sc, "rx_int_delay",
989             "receive interrupt delay in usecs", &sc->rx_int_delay,
990             E1000_REGISTER(hw, E1000_RDTR), em_rx_int_delay_dflt);
991         em_add_int_delay_sysctl(sc, "tx_int_delay",
992             "transmit interrupt delay in usecs", &sc->tx_int_delay,
993             E1000_REGISTER(hw, E1000_TIDV), em_tx_int_delay_dflt);
994         em_add_int_delay_sysctl(sc, "rx_abs_int_delay",
995             "receive interrupt delay limit in usecs",
996             &sc->rx_abs_int_delay,
997             E1000_REGISTER(hw, E1000_RADV),
998             em_rx_abs_int_delay_dflt);
999         em_add_int_delay_sysctl(sc, "tx_abs_int_delay",
1000             "transmit interrupt delay limit in usecs",
1001             &sc->tx_abs_int_delay,
1002             E1000_REGISTER(hw, E1000_TADV),
1003             em_tx_abs_int_delay_dflt);
1004         em_add_int_delay_sysctl(sc, "itr",
1005             "interrupt delay limit in usecs/4",
1006             &sc->tx_itr,
1007             E1000_REGISTER(hw, E1000_ITR),
1008             DEFAULT_ITR);
1009
1010         hw->mac.autoneg = DO_AUTO_NEG;
1011         hw->phy.autoneg_wait_to_complete = false;
1012         hw->phy.autoneg_advertised = AUTONEG_ADV_DEFAULT;
1013
1014         if (hw->mac.type < em_mac_min) {
1015                 e1000_init_script_state_82541(hw, true);
1016                 e1000_set_tbi_compatibility_82543(hw, true);
1017         }
1018         /* Copper options */
1019         if (hw->phy.media_type == e1000_media_type_copper) {
1020                 hw->phy.mdix = AUTO_ALL_MODES;
1021                 hw->phy.disable_polarity_correction = false;
1022                 hw->phy.ms_type = EM_MASTER_SLAVE;
1023         }
1024
1025         /*
1026          * Set the frame limits assuming
1027          * standard ethernet sized frames.
1028          */
1029         scctx->isc_max_frame_size = hw->mac.max_frame_size =
1030             ETHERMTU + ETHER_HDR_LEN + ETHERNET_FCS_SIZE;
1031
1032         /*
1033          * This controls when hardware reports transmit completion
1034          * status.
1035          */
1036         hw->mac.report_tx_early = 1;
1037
1038         /* Allocate multicast array memory. */
1039         sc->mta = malloc(sizeof(u8) * ETHER_ADDR_LEN *
1040             MAX_NUM_MULTICAST_ADDRESSES, M_DEVBUF, M_NOWAIT);
1041         if (sc->mta == NULL) {
1042                 device_printf(dev, "Can not allocate multicast setup array\n");
1043                 error = ENOMEM;
1044                 goto err_late;
1045         }
1046
1047         /* Check SOL/IDER usage */
1048         if (e1000_check_reset_block(hw))
1049                 device_printf(dev, "PHY reset is blocked"
1050                               " due to SOL/IDER session.\n");
1051
1052         /* Sysctl for setting Energy Efficient Ethernet */
1053         hw->dev_spec.ich8lan.eee_disable = eee_setting;
1054         SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "eee_control",
1055             CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT, sc, 0,
1056             em_sysctl_eee, "I", "Disable Energy Efficient Ethernet");
1057
1058         /*
1059         ** Start from a known state, this is
1060         ** important in reading the nvm and
1061         ** mac from that.
1062         */
1063         e1000_reset_hw(hw);
1064
1065         /* Make sure we have a good EEPROM before we read from it */
1066         if (e1000_validate_nvm_checksum(hw) < 0) {
1067                 /*
1068                 ** Some PCI-E parts fail the first check due to
1069                 ** the link being in sleep state, call it again,
1070                 ** if it fails a second time its a real issue.
1071                 */
1072                 if (e1000_validate_nvm_checksum(hw) < 0) {
1073                         device_printf(dev,
1074                             "The EEPROM Checksum Is Not Valid\n");
1075                         error = EIO;
1076                         goto err_late;
1077                 }
1078         }
1079
1080         /* Copy the permanent MAC address out of the EEPROM */
1081         if (e1000_read_mac_addr(hw) < 0) {
1082                 device_printf(dev, "EEPROM read error while reading MAC"
1083                               " address\n");
1084                 error = EIO;
1085                 goto err_late;
1086         }
1087
1088         if (!em_is_valid_ether_addr(hw->mac.addr)) {
1089                 if (sc->vf_ifp) {
1090                         ether_gen_addr(iflib_get_ifp(ctx),
1091                             (struct ether_addr *)hw->mac.addr);
1092                 } else {
1093                         device_printf(dev, "Invalid MAC address\n");
1094                         error = EIO;
1095                         goto err_late;
1096                 }
1097         }
1098
1099         /* Save the EEPROM/NVM versions, must be done under IFLIB_CTX_LOCK */
1100         em_fw_version_locked(ctx);
1101
1102         em_print_fw_version(sc);
1103
1104         /* Disable ULP support */
1105         e1000_disable_ulp_lpt_lp(hw, true);
1106
1107         /*
1108          * Get Wake-on-Lan and Management info for later use
1109          */
1110         em_get_wakeup(ctx);
1111
1112         /* Enable only WOL MAGIC by default */
1113         scctx->isc_capenable &= ~IFCAP_WOL;
1114         if (sc->wol != 0)
1115                 scctx->isc_capenable |= IFCAP_WOL_MAGIC;
1116
1117         iflib_set_mac(ctx, hw->mac.addr);
1118
1119         return (0);
1120
1121 err_late:
1122         em_release_hw_control(sc);
1123 err_pci:
1124         em_free_pci_resources(ctx);
1125         free(sc->mta, M_DEVBUF);
1126
1127         return (error);
1128 }
1129
1130 static int
1131 em_if_attach_post(if_ctx_t ctx)
1132 {
1133         struct e1000_softc *sc = iflib_get_softc(ctx);
1134         struct e1000_hw *hw = &sc->hw;
1135         int error = 0;
1136
1137         /* Setup OS specific network interface */
1138         error = em_setup_interface(ctx);
1139         if (error != 0) {
1140                 device_printf(sc->dev, "Interface setup failed: %d\n", error);
1141                 goto err_late;
1142         }
1143
1144         em_reset(ctx);
1145
1146         /* Initialize statistics */
1147         em_update_stats_counters(sc);
1148         hw->mac.get_link_status = 1;
1149         em_if_update_admin_status(ctx);
1150         em_add_hw_stats(sc);
1151
1152         /* Non-AMT based hardware can now take control from firmware */
1153         if (sc->has_manage && !sc->has_amt)
1154                 em_get_hw_control(sc);
1155
1156         INIT_DEBUGOUT("em_if_attach_post: end");
1157
1158         return (0);
1159
1160 err_late:
1161         /* upon attach_post() error, iflib calls _if_detach() to free resources. */
1162         return (error);
1163 }
1164
1165 /*********************************************************************
1166  *  Device removal routine
1167  *
1168  *  The detach entry point is called when the driver is being removed.
1169  *  This routine stops the adapter and deallocates all the resources
1170  *  that were allocated for driver operation.
1171  *
1172  *  return 0 on success, positive on failure
1173  *********************************************************************/
1174 static int
1175 em_if_detach(if_ctx_t ctx)
1176 {
1177         struct e1000_softc      *sc = iflib_get_softc(ctx);
1178
1179         INIT_DEBUGOUT("em_if_detach: begin");
1180
1181         e1000_phy_hw_reset(&sc->hw);
1182
1183         em_release_manageability(sc);
1184         em_release_hw_control(sc);
1185         em_free_pci_resources(ctx);
1186         free(sc->mta, M_DEVBUF);
1187         sc->mta = NULL;
1188
1189         return (0);
1190 }
1191
1192 /*********************************************************************
1193  *
1194  *  Shutdown entry point
1195  *
1196  **********************************************************************/
1197
1198 static int
1199 em_if_shutdown(if_ctx_t ctx)
1200 {
1201         return em_if_suspend(ctx);
1202 }
1203
1204 /*
1205  * Suspend/resume device methods.
1206  */
1207 static int
1208 em_if_suspend(if_ctx_t ctx)
1209 {
1210         struct e1000_softc *sc = iflib_get_softc(ctx);
1211
1212         em_release_manageability(sc);
1213         em_release_hw_control(sc);
1214         em_enable_wakeup(ctx);
1215         return (0);
1216 }
1217
1218 static int
1219 em_if_resume(if_ctx_t ctx)
1220 {
1221         struct e1000_softc *sc = iflib_get_softc(ctx);
1222
1223         if (sc->hw.mac.type == e1000_pch2lan)
1224                 e1000_resume_workarounds_pchlan(&sc->hw);
1225         em_if_init(ctx);
1226         em_init_manageability(sc);
1227
1228         return(0);
1229 }
1230
1231 static int
1232 em_if_mtu_set(if_ctx_t ctx, uint32_t mtu)
1233 {
1234         int max_frame_size;
1235         struct e1000_softc *sc = iflib_get_softc(ctx);
1236         if_softc_ctx_t scctx = iflib_get_softc_ctx(ctx);
1237
1238         IOCTL_DEBUGOUT("ioctl rcv'd: SIOCSIFMTU (Set Interface MTU)");
1239
1240         switch (sc->hw.mac.type) {
1241         case e1000_82571:
1242         case e1000_82572:
1243         case e1000_ich9lan:
1244         case e1000_ich10lan:
1245         case e1000_pch2lan:
1246         case e1000_pch_lpt:
1247         case e1000_pch_spt:
1248         case e1000_pch_cnp:
1249         case e1000_pch_tgp:
1250         case e1000_pch_adp:
1251         case e1000_pch_mtp:
1252         case e1000_82574:
1253         case e1000_82583:
1254         case e1000_80003es2lan:
1255                 /* 9K Jumbo Frame size */
1256                 max_frame_size = 9234;
1257                 break;
1258         case e1000_pchlan:
1259                 max_frame_size = 4096;
1260                 break;
1261         case e1000_82542:
1262         case e1000_ich8lan:
1263                 /* Adapters that do not support jumbo frames */
1264                 max_frame_size = ETHER_MAX_LEN;
1265                 break;
1266         default:
1267                 if (sc->hw.mac.type >= igb_mac_min)
1268                         max_frame_size = 9234;
1269                 else /* lem */
1270                         max_frame_size = MAX_JUMBO_FRAME_SIZE;
1271         }
1272         if (mtu > max_frame_size - ETHER_HDR_LEN - ETHER_CRC_LEN) {
1273                 return (EINVAL);
1274         }
1275
1276         scctx->isc_max_frame_size = sc->hw.mac.max_frame_size =
1277             mtu + ETHER_HDR_LEN + ETHER_CRC_LEN;
1278         return (0);
1279 }
1280
1281 /*********************************************************************
1282  *  Init entry point
1283  *
1284  *  This routine is used in two ways. It is used by the stack as
1285  *  init entry point in network interface structure. It is also used
1286  *  by the driver as a hw/sw initialization routine to get to a
1287  *  consistent state.
1288  *
1289  **********************************************************************/
1290 static void
1291 em_if_init(if_ctx_t ctx)
1292 {
1293         struct e1000_softc *sc = iflib_get_softc(ctx);
1294         if_softc_ctx_t scctx = sc->shared;
1295         struct ifnet *ifp = iflib_get_ifp(ctx);
1296         struct em_tx_queue *tx_que;
1297         int i;
1298
1299         INIT_DEBUGOUT("em_if_init: begin");
1300
1301         /* Get the latest mac address, User can use a LAA */
1302         bcopy(if_getlladdr(ifp), sc->hw.mac.addr,
1303             ETHER_ADDR_LEN);
1304
1305         /* Put the address into the Receive Address Array */
1306         e1000_rar_set(&sc->hw, sc->hw.mac.addr, 0);
1307
1308         /*
1309          * With the 82571 adapter, RAR[0] may be overwritten
1310          * when the other port is reset, we make a duplicate
1311          * in RAR[14] for that eventuality, this assures
1312          * the interface continues to function.
1313          */
1314         if (sc->hw.mac.type == e1000_82571) {
1315                 e1000_set_laa_state_82571(&sc->hw, true);
1316                 e1000_rar_set(&sc->hw, sc->hw.mac.addr,
1317                     E1000_RAR_ENTRIES - 1);
1318         }
1319
1320
1321         /* Initialize the hardware */
1322         em_reset(ctx);
1323         em_if_update_admin_status(ctx);
1324
1325         for (i = 0, tx_que = sc->tx_queues; i < sc->tx_num_queues; i++, tx_que++) {
1326                 struct tx_ring *txr = &tx_que->txr;
1327
1328                 txr->tx_rs_cidx = txr->tx_rs_pidx;
1329
1330                 /* Initialize the last processed descriptor to be the end of
1331                  * the ring, rather than the start, so that we avoid an
1332                  * off-by-one error when calculating how many descriptors are
1333                  * done in the credits_update function.
1334                  */
1335                 txr->tx_cidx_processed = scctx->isc_ntxd[0] - 1;
1336         }
1337
1338         /* Setup VLAN support, basic and offload if available */
1339         E1000_WRITE_REG(&sc->hw, E1000_VET, ETHERTYPE_VLAN);
1340
1341         /* Clear bad data from Rx FIFOs */
1342         if (sc->hw.mac.type >= igb_mac_min)
1343                 e1000_rx_fifo_flush_base(&sc->hw);
1344
1345         /* Configure for OS presence */
1346         em_init_manageability(sc);
1347
1348         /* Prepare transmit descriptors and buffers */
1349         em_initialize_transmit_unit(ctx);
1350
1351         /* Setup Multicast table */
1352         em_if_multi_set(ctx);
1353
1354         sc->rx_mbuf_sz = iflib_get_rx_mbuf_sz(ctx);
1355         em_initialize_receive_unit(ctx);
1356
1357         /* Set up VLAN support and filter */
1358         em_setup_vlan_hw_support(ctx);
1359
1360         /* Don't lose promiscuous settings */
1361         em_if_set_promisc(ctx, if_getflags(ifp));
1362         e1000_clear_hw_cntrs_base_generic(&sc->hw);
1363
1364         /* MSI-X configuration for 82574 */
1365         if (sc->hw.mac.type == e1000_82574) {
1366                 int tmp = E1000_READ_REG(&sc->hw, E1000_CTRL_EXT);
1367
1368                 tmp |= E1000_CTRL_EXT_PBA_CLR;
1369                 E1000_WRITE_REG(&sc->hw, E1000_CTRL_EXT, tmp);
1370                 /* Set the IVAR - interrupt vector routing. */
1371                 E1000_WRITE_REG(&sc->hw, E1000_IVAR, sc->ivars);
1372         } else if (sc->intr_type == IFLIB_INTR_MSIX) /* Set up queue routing */
1373                 igb_configure_queues(sc);
1374
1375         /* this clears any pending interrupts */
1376         E1000_READ_REG(&sc->hw, E1000_ICR);
1377         E1000_WRITE_REG(&sc->hw, E1000_ICS, E1000_ICS_LSC);
1378
1379         /* AMT based hardware can now take control from firmware */
1380         if (sc->has_manage && sc->has_amt)
1381                 em_get_hw_control(sc);
1382
1383         /* Set Energy Efficient Ethernet */
1384         if (sc->hw.mac.type >= igb_mac_min &&
1385             sc->hw.phy.media_type == e1000_media_type_copper) {
1386                 if (sc->hw.mac.type == e1000_i354)
1387                         e1000_set_eee_i354(&sc->hw, true, true);
1388                 else
1389                         e1000_set_eee_i350(&sc->hw, true, true);
1390         }
1391 }
1392
1393 /*********************************************************************
1394  *
1395  *  Fast Legacy/MSI Combined Interrupt Service routine
1396  *
1397  *********************************************************************/
1398 int
1399 em_intr(void *arg)
1400 {
1401         struct e1000_softc *sc = arg;
1402         if_ctx_t ctx = sc->ctx;
1403         u32 reg_icr;
1404
1405         reg_icr = E1000_READ_REG(&sc->hw, E1000_ICR);
1406
1407         /* Hot eject? */
1408         if (reg_icr == 0xffffffff)
1409                 return FILTER_STRAY;
1410
1411         /* Definitely not our interrupt. */
1412         if (reg_icr == 0x0)
1413                 return FILTER_STRAY;
1414
1415         /*
1416          * Starting with the 82571 chip, bit 31 should be used to
1417          * determine whether the interrupt belongs to us.
1418          */
1419         if (sc->hw.mac.type >= e1000_82571 &&
1420             (reg_icr & E1000_ICR_INT_ASSERTED) == 0)
1421                 return FILTER_STRAY;
1422
1423         /*
1424          * Only MSI-X interrupts have one-shot behavior by taking advantage
1425          * of the EIAC register.  Thus, explicitly disable interrupts.  This
1426          * also works around the MSI message reordering errata on certain
1427          * systems.
1428          */
1429         IFDI_INTR_DISABLE(ctx);
1430
1431         /* Link status change */
1432         if (reg_icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC))
1433                 em_handle_link(ctx);
1434
1435         if (reg_icr & E1000_ICR_RXO)
1436                 sc->rx_overruns++;
1437
1438         return (FILTER_SCHEDULE_THREAD);
1439 }
1440
1441 static int
1442 em_if_rx_queue_intr_enable(if_ctx_t ctx, uint16_t rxqid)
1443 {
1444         struct e1000_softc *sc = iflib_get_softc(ctx);
1445         struct em_rx_queue *rxq = &sc->rx_queues[rxqid];
1446
1447         E1000_WRITE_REG(&sc->hw, E1000_IMS, rxq->eims);
1448         return (0);
1449 }
1450
1451 static int
1452 em_if_tx_queue_intr_enable(if_ctx_t ctx, uint16_t txqid)
1453 {
1454         struct e1000_softc *sc = iflib_get_softc(ctx);
1455         struct em_tx_queue *txq = &sc->tx_queues[txqid];
1456
1457         E1000_WRITE_REG(&sc->hw, E1000_IMS, txq->eims);
1458         return (0);
1459 }
1460
1461 static int
1462 igb_if_rx_queue_intr_enable(if_ctx_t ctx, uint16_t rxqid)
1463 {
1464         struct e1000_softc *sc = iflib_get_softc(ctx);
1465         struct em_rx_queue *rxq = &sc->rx_queues[rxqid];
1466
1467         E1000_WRITE_REG(&sc->hw, E1000_EIMS, rxq->eims);
1468         return (0);
1469 }
1470
1471 static int
1472 igb_if_tx_queue_intr_enable(if_ctx_t ctx, uint16_t txqid)
1473 {
1474         struct e1000_softc *sc = iflib_get_softc(ctx);
1475         struct em_tx_queue *txq = &sc->tx_queues[txqid];
1476
1477         E1000_WRITE_REG(&sc->hw, E1000_EIMS, txq->eims);
1478         return (0);
1479 }
1480
1481 /*********************************************************************
1482  *
1483  *  MSI-X RX Interrupt Service routine
1484  *
1485  **********************************************************************/
1486 static int
1487 em_msix_que(void *arg)
1488 {
1489         struct em_rx_queue *que = arg;
1490
1491         ++que->irqs;
1492
1493         return (FILTER_SCHEDULE_THREAD);
1494 }
1495
1496 /*********************************************************************
1497  *
1498  *  MSI-X Link Fast Interrupt Service routine
1499  *
1500  **********************************************************************/
1501 static int
1502 em_msix_link(void *arg)
1503 {
1504         struct e1000_softc *sc = arg;
1505         u32 reg_icr;
1506
1507         ++sc->link_irq;
1508         MPASS(sc->hw.back != NULL);
1509         reg_icr = E1000_READ_REG(&sc->hw, E1000_ICR);
1510
1511         if (reg_icr & E1000_ICR_RXO)
1512                 sc->rx_overruns++;
1513
1514         if (reg_icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC))
1515                 em_handle_link(sc->ctx);
1516
1517         /* Re-arm unconditionally */
1518         if (sc->hw.mac.type >= igb_mac_min) {
1519                 E1000_WRITE_REG(&sc->hw, E1000_IMS, E1000_IMS_LSC);
1520                 E1000_WRITE_REG(&sc->hw, E1000_EIMS, sc->link_mask);
1521         } else if (sc->hw.mac.type == e1000_82574) {
1522                 E1000_WRITE_REG(&sc->hw, E1000_IMS, E1000_IMS_LSC |
1523                     E1000_IMS_OTHER);
1524                 /*
1525                  * Because we must read the ICR for this interrupt it may
1526                  * clear other causes using autoclear, for this reason we
1527                  * simply create a soft interrupt for all these vectors.
1528                  */
1529                 if (reg_icr)
1530                         E1000_WRITE_REG(&sc->hw, E1000_ICS, sc->ims);
1531         } else
1532                 E1000_WRITE_REG(&sc->hw, E1000_IMS, E1000_IMS_LSC);
1533
1534         return (FILTER_HANDLED);
1535 }
1536
1537 static void
1538 em_handle_link(void *context)
1539 {
1540         if_ctx_t ctx = context;
1541         struct e1000_softc *sc = iflib_get_softc(ctx);
1542
1543         sc->hw.mac.get_link_status = 1;
1544         iflib_admin_intr_deferred(ctx);
1545 }
1546
1547 /*********************************************************************
1548  *
1549  *  Media Ioctl callback
1550  *
1551  *  This routine is called whenever the user queries the status of
1552  *  the interface using ifconfig.
1553  *
1554  **********************************************************************/
1555 static void
1556 em_if_media_status(if_ctx_t ctx, struct ifmediareq *ifmr)
1557 {
1558         struct e1000_softc *sc = iflib_get_softc(ctx);
1559         u_char fiber_type = IFM_1000_SX;
1560
1561         INIT_DEBUGOUT("em_if_media_status: begin");
1562
1563         iflib_admin_intr_deferred(ctx);
1564
1565         ifmr->ifm_status = IFM_AVALID;
1566         ifmr->ifm_active = IFM_ETHER;
1567
1568         if (!sc->link_active) {
1569                 return;
1570         }
1571
1572         ifmr->ifm_status |= IFM_ACTIVE;
1573
1574         if ((sc->hw.phy.media_type == e1000_media_type_fiber) ||
1575             (sc->hw.phy.media_type == e1000_media_type_internal_serdes)) {
1576                 if (sc->hw.mac.type == e1000_82545)
1577                         fiber_type = IFM_1000_LX;
1578                 ifmr->ifm_active |= fiber_type | IFM_FDX;
1579         } else {
1580                 switch (sc->link_speed) {
1581                 case 10:
1582                         ifmr->ifm_active |= IFM_10_T;
1583                         break;
1584                 case 100:
1585                         ifmr->ifm_active |= IFM_100_TX;
1586                         break;
1587                 case 1000:
1588                         ifmr->ifm_active |= IFM_1000_T;
1589                         break;
1590                 }
1591                 if (sc->link_duplex == FULL_DUPLEX)
1592                         ifmr->ifm_active |= IFM_FDX;
1593                 else
1594                         ifmr->ifm_active |= IFM_HDX;
1595         }
1596 }
1597
1598 /*********************************************************************
1599  *
1600  *  Media Ioctl callback
1601  *
1602  *  This routine is called when the user changes speed/duplex using
1603  *  media/mediopt option with ifconfig.
1604  *
1605  **********************************************************************/
1606 static int
1607 em_if_media_change(if_ctx_t ctx)
1608 {
1609         struct e1000_softc *sc = iflib_get_softc(ctx);
1610         struct ifmedia *ifm = iflib_get_media(ctx);
1611
1612         INIT_DEBUGOUT("em_if_media_change: begin");
1613
1614         if (IFM_TYPE(ifm->ifm_media) != IFM_ETHER)
1615                 return (EINVAL);
1616
1617         switch (IFM_SUBTYPE(ifm->ifm_media)) {
1618         case IFM_AUTO:
1619                 sc->hw.mac.autoneg = DO_AUTO_NEG;
1620                 sc->hw.phy.autoneg_advertised = AUTONEG_ADV_DEFAULT;
1621                 break;
1622         case IFM_1000_LX:
1623         case IFM_1000_SX:
1624         case IFM_1000_T:
1625                 sc->hw.mac.autoneg = DO_AUTO_NEG;
1626                 sc->hw.phy.autoneg_advertised = ADVERTISE_1000_FULL;
1627                 break;
1628         case IFM_100_TX:
1629                 sc->hw.mac.autoneg = DO_AUTO_NEG;
1630                 if ((ifm->ifm_media & IFM_GMASK) == IFM_FDX) {
1631                         sc->hw.phy.autoneg_advertised = ADVERTISE_100_FULL;
1632                         sc->hw.mac.forced_speed_duplex = ADVERTISE_100_FULL;
1633                 } else {
1634                         sc->hw.phy.autoneg_advertised = ADVERTISE_100_HALF;
1635                         sc->hw.mac.forced_speed_duplex = ADVERTISE_100_HALF;
1636                 }
1637                 break;
1638         case IFM_10_T:
1639                 sc->hw.mac.autoneg = DO_AUTO_NEG;
1640                 if ((ifm->ifm_media & IFM_GMASK) == IFM_FDX) {
1641                         sc->hw.phy.autoneg_advertised = ADVERTISE_10_FULL;
1642                         sc->hw.mac.forced_speed_duplex = ADVERTISE_10_FULL;
1643                 } else {
1644                         sc->hw.phy.autoneg_advertised = ADVERTISE_10_HALF;
1645                         sc->hw.mac.forced_speed_duplex = ADVERTISE_10_HALF;
1646                 }
1647                 break;
1648         default:
1649                 device_printf(sc->dev, "Unsupported media type\n");
1650         }
1651
1652         em_if_init(ctx);
1653
1654         return (0);
1655 }
1656
1657 static int
1658 em_if_set_promisc(if_ctx_t ctx, int flags)
1659 {
1660         struct e1000_softc *sc = iflib_get_softc(ctx);
1661         struct ifnet *ifp = iflib_get_ifp(ctx);
1662         u32 reg_rctl;
1663         int mcnt = 0;
1664
1665         reg_rctl = E1000_READ_REG(&sc->hw, E1000_RCTL);
1666         reg_rctl &= ~(E1000_RCTL_SBP | E1000_RCTL_UPE);
1667         if (flags & IFF_ALLMULTI)
1668                 mcnt = MAX_NUM_MULTICAST_ADDRESSES;
1669         else
1670                 mcnt = min(if_llmaddr_count(ifp), MAX_NUM_MULTICAST_ADDRESSES);
1671
1672         if (mcnt < MAX_NUM_MULTICAST_ADDRESSES)
1673                 reg_rctl &= (~E1000_RCTL_MPE);
1674         E1000_WRITE_REG(&sc->hw, E1000_RCTL, reg_rctl);
1675
1676         if (flags & IFF_PROMISC) {
1677                 reg_rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
1678                 em_if_vlan_filter_disable(sc);
1679                 /* Turn this on if you want to see bad packets */
1680                 if (em_debug_sbp)
1681                         reg_rctl |= E1000_RCTL_SBP;
1682                 E1000_WRITE_REG(&sc->hw, E1000_RCTL, reg_rctl);
1683         } else {
1684                 if (flags & IFF_ALLMULTI) {
1685                         reg_rctl |= E1000_RCTL_MPE;
1686                         reg_rctl &= ~E1000_RCTL_UPE;
1687                         E1000_WRITE_REG(&sc->hw, E1000_RCTL, reg_rctl);
1688                 }
1689                 if (em_if_vlan_filter_used(ctx))
1690                         em_if_vlan_filter_enable(sc);
1691         }
1692         return (0);
1693 }
1694
1695 static u_int
1696 em_copy_maddr(void *arg, struct sockaddr_dl *sdl, u_int idx)
1697 {
1698         u8 *mta = arg;
1699
1700         if (idx == MAX_NUM_MULTICAST_ADDRESSES)
1701                 return (0);
1702
1703         bcopy(LLADDR(sdl), &mta[idx * ETHER_ADDR_LEN], ETHER_ADDR_LEN);
1704
1705         return (1);
1706 }
1707
1708 /*********************************************************************
1709  *  Multicast Update
1710  *
1711  *  This routine is called whenever multicast address list is updated.
1712  *
1713  **********************************************************************/
1714 static void
1715 em_if_multi_set(if_ctx_t ctx)
1716 {
1717         struct e1000_softc *sc = iflib_get_softc(ctx);
1718         struct ifnet *ifp = iflib_get_ifp(ctx);
1719         u8  *mta; /* Multicast array memory */
1720         u32 reg_rctl = 0;
1721         int mcnt = 0;
1722
1723         IOCTL_DEBUGOUT("em_set_multi: begin");
1724
1725         mta = sc->mta;
1726         bzero(mta, sizeof(u8) * ETHER_ADDR_LEN * MAX_NUM_MULTICAST_ADDRESSES);
1727
1728         if (sc->hw.mac.type == e1000_82542 &&
1729             sc->hw.revision_id == E1000_REVISION_2) {
1730                 reg_rctl = E1000_READ_REG(&sc->hw, E1000_RCTL);
1731                 if (sc->hw.bus.pci_cmd_word & CMD_MEM_WRT_INVALIDATE)
1732                         e1000_pci_clear_mwi(&sc->hw);
1733                 reg_rctl |= E1000_RCTL_RST;
1734                 E1000_WRITE_REG(&sc->hw, E1000_RCTL, reg_rctl);
1735                 msec_delay(5);
1736         }
1737
1738         mcnt = if_foreach_llmaddr(ifp, em_copy_maddr, mta);
1739
1740         if (mcnt < MAX_NUM_MULTICAST_ADDRESSES)
1741                 e1000_update_mc_addr_list(&sc->hw, mta, mcnt);
1742
1743         reg_rctl = E1000_READ_REG(&sc->hw, E1000_RCTL);
1744
1745         if (if_getflags(ifp) & IFF_PROMISC)
1746                 reg_rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
1747         else if (mcnt >= MAX_NUM_MULTICAST_ADDRESSES ||
1748             if_getflags(ifp) & IFF_ALLMULTI) {
1749                 reg_rctl |= E1000_RCTL_MPE;
1750                 reg_rctl &= ~E1000_RCTL_UPE;
1751         } else
1752                 reg_rctl &= ~(E1000_RCTL_UPE | E1000_RCTL_MPE);
1753
1754         E1000_WRITE_REG(&sc->hw, E1000_RCTL, reg_rctl);
1755
1756         if (sc->hw.mac.type == e1000_82542 &&
1757             sc->hw.revision_id == E1000_REVISION_2) {
1758                 reg_rctl = E1000_READ_REG(&sc->hw, E1000_RCTL);
1759                 reg_rctl &= ~E1000_RCTL_RST;
1760                 E1000_WRITE_REG(&sc->hw, E1000_RCTL, reg_rctl);
1761                 msec_delay(5);
1762                 if (sc->hw.bus.pci_cmd_word & CMD_MEM_WRT_INVALIDATE)
1763                         e1000_pci_set_mwi(&sc->hw);
1764         }
1765 }
1766
1767 /*********************************************************************
1768  *  Timer routine
1769  *
1770  *  This routine schedules em_if_update_admin_status() to check for
1771  *  link status and to gather statistics as well as to perform some
1772  *  controller-specific hardware patting.
1773  *
1774  **********************************************************************/
1775 static void
1776 em_if_timer(if_ctx_t ctx, uint16_t qid)
1777 {
1778
1779         if (qid != 0)
1780                 return;
1781
1782         iflib_admin_intr_deferred(ctx);
1783 }
1784
1785 static void
1786 em_if_update_admin_status(if_ctx_t ctx)
1787 {
1788         struct e1000_softc *sc = iflib_get_softc(ctx);
1789         struct e1000_hw *hw = &sc->hw;
1790         device_t dev = iflib_get_dev(ctx);
1791         u32 link_check, thstat, ctrl;
1792
1793         link_check = thstat = ctrl = 0;
1794         /* Get the cached link value or read phy for real */
1795         switch (hw->phy.media_type) {
1796         case e1000_media_type_copper:
1797                 if (hw->mac.get_link_status) {
1798                         if (hw->mac.type == e1000_pch_spt)
1799                                 msec_delay(50);
1800                         /* Do the work to read phy */
1801                         e1000_check_for_link(hw);
1802                         link_check = !hw->mac.get_link_status;
1803                         if (link_check) /* ESB2 fix */
1804                                 e1000_cfg_on_link_up(hw);
1805                 } else {
1806                         link_check = true;
1807                 }
1808                 break;
1809         case e1000_media_type_fiber:
1810                 e1000_check_for_link(hw);
1811                 link_check = (E1000_READ_REG(hw, E1000_STATUS) &
1812                             E1000_STATUS_LU);
1813                 break;
1814         case e1000_media_type_internal_serdes:
1815                 e1000_check_for_link(hw);
1816                 link_check = hw->mac.serdes_has_link;
1817                 break;
1818         /* VF device is type_unknown */
1819         case e1000_media_type_unknown:
1820                 e1000_check_for_link(hw);
1821                 link_check = !hw->mac.get_link_status;
1822                 /* FALLTHROUGH */
1823         default:
1824                 break;
1825         }
1826
1827         /* Check for thermal downshift or shutdown */
1828         if (hw->mac.type == e1000_i350) {
1829                 thstat = E1000_READ_REG(hw, E1000_THSTAT);
1830                 ctrl = E1000_READ_REG(hw, E1000_CTRL_EXT);
1831         }
1832
1833         /* Now check for a transition */
1834         if (link_check && (sc->link_active == 0)) {
1835                 e1000_get_speed_and_duplex(hw, &sc->link_speed,
1836                     &sc->link_duplex);
1837                 /* Check if we must disable SPEED_MODE bit on PCI-E */
1838                 if ((sc->link_speed != SPEED_1000) &&
1839                     ((hw->mac.type == e1000_82571) ||
1840                     (hw->mac.type == e1000_82572))) {
1841                         int tarc0;
1842                         tarc0 = E1000_READ_REG(hw, E1000_TARC(0));
1843                         tarc0 &= ~TARC_SPEED_MODE_BIT;
1844                         E1000_WRITE_REG(hw, E1000_TARC(0), tarc0);
1845                 }
1846                 if (bootverbose)
1847                         device_printf(dev, "Link is up %d Mbps %s\n",
1848                             sc->link_speed,
1849                             ((sc->link_duplex == FULL_DUPLEX) ?
1850                             "Full Duplex" : "Half Duplex"));
1851                 sc->link_active = 1;
1852                 sc->smartspeed = 0;
1853                 if ((ctrl & E1000_CTRL_EXT_LINK_MODE_MASK) ==
1854                     E1000_CTRL_EXT_LINK_MODE_GMII &&
1855                     (thstat & E1000_THSTAT_LINK_THROTTLE))
1856                         device_printf(dev, "Link: thermal downshift\n");
1857                 /* Delay Link Up for Phy update */
1858                 if (((hw->mac.type == e1000_i210) ||
1859                     (hw->mac.type == e1000_i211)) &&
1860                     (hw->phy.id == I210_I_PHY_ID))
1861                         msec_delay(I210_LINK_DELAY);
1862                 /* Reset if the media type changed. */
1863                 if (hw->dev_spec._82575.media_changed &&
1864                     hw->mac.type >= igb_mac_min) {
1865                         hw->dev_spec._82575.media_changed = false;
1866                         sc->flags |= IGB_MEDIA_RESET;
1867                         em_reset(ctx);
1868                 }
1869                 iflib_link_state_change(ctx, LINK_STATE_UP,
1870                     IF_Mbps(sc->link_speed));
1871         } else if (!link_check && (sc->link_active == 1)) {
1872                 sc->link_speed = 0;
1873                 sc->link_duplex = 0;
1874                 sc->link_active = 0;
1875                 iflib_link_state_change(ctx, LINK_STATE_DOWN, 0);
1876         }
1877         em_update_stats_counters(sc);
1878
1879         /* Reset LAA into RAR[0] on 82571 */
1880         if (hw->mac.type == e1000_82571 && e1000_get_laa_state_82571(hw))
1881                 e1000_rar_set(hw, hw->mac.addr, 0);
1882
1883         if (hw->mac.type < em_mac_min)
1884                 lem_smartspeed(sc);
1885 }
1886
1887 static void
1888 em_if_watchdog_reset(if_ctx_t ctx)
1889 {
1890         struct e1000_softc *sc = iflib_get_softc(ctx);
1891
1892         /*
1893          * Just count the event; iflib(4) will already trigger a
1894          * sufficient reset of the controller.
1895          */
1896         sc->watchdog_events++;
1897 }
1898
1899 /*********************************************************************
1900  *
1901  *  This routine disables all traffic on the adapter by issuing a
1902  *  global reset on the MAC.
1903  *
1904  **********************************************************************/
1905 static void
1906 em_if_stop(if_ctx_t ctx)
1907 {
1908         struct e1000_softc *sc = iflib_get_softc(ctx);
1909
1910         INIT_DEBUGOUT("em_if_stop: begin");
1911
1912         e1000_reset_hw(&sc->hw);
1913         if (sc->hw.mac.type >= e1000_82544)
1914                 E1000_WRITE_REG(&sc->hw, E1000_WUFC, 0);
1915
1916         e1000_led_off(&sc->hw);
1917         e1000_cleanup_led(&sc->hw);
1918 }
1919
1920 /*********************************************************************
1921  *
1922  *  Determine hardware revision.
1923  *
1924  **********************************************************************/
1925 static void
1926 em_identify_hardware(if_ctx_t ctx)
1927 {
1928         device_t dev = iflib_get_dev(ctx);
1929         struct e1000_softc *sc = iflib_get_softc(ctx);
1930
1931         /* Make sure our PCI config space has the necessary stuff set */
1932         sc->hw.bus.pci_cmd_word = pci_read_config(dev, PCIR_COMMAND, 2);
1933
1934         /* Save off the information about this board */
1935         sc->hw.vendor_id = pci_get_vendor(dev);
1936         sc->hw.device_id = pci_get_device(dev);
1937         sc->hw.revision_id = pci_read_config(dev, PCIR_REVID, 1);
1938         sc->hw.subsystem_vendor_id =
1939             pci_read_config(dev, PCIR_SUBVEND_0, 2);
1940         sc->hw.subsystem_device_id =
1941             pci_read_config(dev, PCIR_SUBDEV_0, 2);
1942
1943         /* Do Shared Code Init and Setup */
1944         if (e1000_set_mac_type(&sc->hw)) {
1945                 device_printf(dev, "Setup init failure\n");
1946                 return;
1947         }
1948
1949         /* Are we a VF device? */
1950         if ((sc->hw.mac.type == e1000_vfadapt) ||
1951             (sc->hw.mac.type == e1000_vfadapt_i350))
1952                 sc->vf_ifp = 1;
1953         else
1954                 sc->vf_ifp = 0;
1955 }
1956
1957 static int
1958 em_allocate_pci_resources(if_ctx_t ctx)
1959 {
1960         struct e1000_softc *sc = iflib_get_softc(ctx);
1961         device_t dev = iflib_get_dev(ctx);
1962         int rid, val;
1963
1964         rid = PCIR_BAR(0);
1965         sc->memory = bus_alloc_resource_any(dev, SYS_RES_MEMORY,
1966             &rid, RF_ACTIVE);
1967         if (sc->memory == NULL) {
1968                 device_printf(dev, "Unable to allocate bus resource: memory\n");
1969                 return (ENXIO);
1970         }
1971         sc->osdep.mem_bus_space_tag = rman_get_bustag(sc->memory);
1972         sc->osdep.mem_bus_space_handle =
1973             rman_get_bushandle(sc->memory);
1974         sc->hw.hw_addr = (u8 *)&sc->osdep.mem_bus_space_handle;
1975
1976         /* Only older adapters use IO mapping */
1977         if (sc->hw.mac.type < em_mac_min &&
1978             sc->hw.mac.type > e1000_82543) {
1979                 /* Figure our where our IO BAR is ? */
1980                 for (rid = PCIR_BAR(0); rid < PCIR_CIS;) {
1981                         val = pci_read_config(dev, rid, 4);
1982                         if (EM_BAR_TYPE(val) == EM_BAR_TYPE_IO) {
1983                                 break;
1984                         }
1985                         rid += 4;
1986                         /* check for 64bit BAR */
1987                         if (EM_BAR_MEM_TYPE(val) == EM_BAR_MEM_TYPE_64BIT)
1988                                 rid += 4;
1989                 }
1990                 if (rid >= PCIR_CIS) {
1991                         device_printf(dev, "Unable to locate IO BAR\n");
1992                         return (ENXIO);
1993                 }
1994                 sc->ioport = bus_alloc_resource_any(dev, SYS_RES_IOPORT,
1995                     &rid, RF_ACTIVE);
1996                 if (sc->ioport == NULL) {
1997                         device_printf(dev, "Unable to allocate bus resource: "
1998                             "ioport\n");
1999                         return (ENXIO);
2000                 }
2001                 sc->hw.io_base = 0;
2002                 sc->osdep.io_bus_space_tag =
2003                     rman_get_bustag(sc->ioport);
2004                 sc->osdep.io_bus_space_handle =
2005                     rman_get_bushandle(sc->ioport);
2006         }
2007
2008         sc->hw.back = &sc->osdep;
2009
2010         return (0);
2011 }
2012
2013 /*********************************************************************
2014  *
2015  *  Set up the MSI-X Interrupt handlers
2016  *
2017  **********************************************************************/
2018 static int
2019 em_if_msix_intr_assign(if_ctx_t ctx, int msix)
2020 {
2021         struct e1000_softc *sc = iflib_get_softc(ctx);
2022         struct em_rx_queue *rx_que = sc->rx_queues;
2023         struct em_tx_queue *tx_que = sc->tx_queues;
2024         int error, rid, i, vector = 0, rx_vectors;
2025         char buf[16];
2026
2027         /* First set up ring resources */
2028         for (i = 0; i < sc->rx_num_queues; i++, rx_que++, vector++) {
2029                 rid = vector + 1;
2030                 snprintf(buf, sizeof(buf), "rxq%d", i);
2031                 error = iflib_irq_alloc_generic(ctx, &rx_que->que_irq, rid, IFLIB_INTR_RXTX, em_msix_que, rx_que, rx_que->me, buf);
2032                 if (error) {
2033                         device_printf(iflib_get_dev(ctx), "Failed to allocate que int %d err: %d", i, error);
2034                         sc->rx_num_queues = i + 1;
2035                         goto fail;
2036                 }
2037
2038                 rx_que->msix =  vector;
2039
2040                 /*
2041                  * Set the bit to enable interrupt
2042                  * in E1000_IMS -- bits 20 and 21
2043                  * are for RX0 and RX1, note this has
2044                  * NOTHING to do with the MSI-X vector
2045                  */
2046                 if (sc->hw.mac.type == e1000_82574) {
2047                         rx_que->eims = 1 << (20 + i);
2048                         sc->ims |= rx_que->eims;
2049                         sc->ivars |= (8 | rx_que->msix) << (i * 4);
2050                 } else if (sc->hw.mac.type == e1000_82575)
2051                         rx_que->eims = E1000_EICR_TX_QUEUE0 << vector;
2052                 else
2053                         rx_que->eims = 1 << vector;
2054         }
2055         rx_vectors = vector;
2056
2057         vector = 0;
2058         for (i = 0; i < sc->tx_num_queues; i++, tx_que++, vector++) {
2059                 snprintf(buf, sizeof(buf), "txq%d", i);
2060                 tx_que = &sc->tx_queues[i];
2061                 iflib_softirq_alloc_generic(ctx,
2062                     &sc->rx_queues[i % sc->rx_num_queues].que_irq,
2063                     IFLIB_INTR_TX, tx_que, tx_que->me, buf);
2064
2065                 tx_que->msix = (vector % sc->rx_num_queues);
2066
2067                 /*
2068                  * Set the bit to enable interrupt
2069                  * in E1000_IMS -- bits 22 and 23
2070                  * are for TX0 and TX1, note this has
2071                  * NOTHING to do with the MSI-X vector
2072                  */
2073                 if (sc->hw.mac.type == e1000_82574) {
2074                         tx_que->eims = 1 << (22 + i);
2075                         sc->ims |= tx_que->eims;
2076                         sc->ivars |= (8 | tx_que->msix) << (8 + (i * 4));
2077                 } else if (sc->hw.mac.type == e1000_82575) {
2078                         tx_que->eims = E1000_EICR_TX_QUEUE0 << i;
2079                 } else {
2080                         tx_que->eims = 1 << i;
2081                 }
2082         }
2083
2084         /* Link interrupt */
2085         rid = rx_vectors + 1;
2086         error = iflib_irq_alloc_generic(ctx, &sc->irq, rid, IFLIB_INTR_ADMIN, em_msix_link, sc, 0, "aq");
2087
2088         if (error) {
2089                 device_printf(iflib_get_dev(ctx), "Failed to register admin handler");
2090                 goto fail;
2091         }
2092         sc->linkvec = rx_vectors;
2093         if (sc->hw.mac.type < igb_mac_min) {
2094                 sc->ivars |=  (8 | rx_vectors) << 16;
2095                 sc->ivars |= 0x80000000;
2096                 /* Enable the "Other" interrupt type for link status change */
2097                 sc->ims |= E1000_IMS_OTHER;
2098         }
2099
2100         return (0);
2101 fail:
2102         iflib_irq_free(ctx, &sc->irq);
2103         rx_que = sc->rx_queues;
2104         for (int i = 0; i < sc->rx_num_queues; i++, rx_que++)
2105                 iflib_irq_free(ctx, &rx_que->que_irq);
2106         return (error);
2107 }
2108
2109 static void
2110 igb_configure_queues(struct e1000_softc *sc)
2111 {
2112         struct e1000_hw *hw = &sc->hw;
2113         struct em_rx_queue *rx_que;
2114         struct em_tx_queue *tx_que;
2115         u32 tmp, ivar = 0, newitr = 0;
2116
2117         /* First turn on RSS capability */
2118         if (hw->mac.type != e1000_82575)
2119                 E1000_WRITE_REG(hw, E1000_GPIE,
2120                     E1000_GPIE_MSIX_MODE | E1000_GPIE_EIAME |
2121                     E1000_GPIE_PBA | E1000_GPIE_NSICR);
2122
2123         /* Turn on MSI-X */
2124         switch (hw->mac.type) {
2125         case e1000_82580:
2126         case e1000_i350:
2127         case e1000_i354:
2128         case e1000_i210:
2129         case e1000_i211:
2130         case e1000_vfadapt:
2131         case e1000_vfadapt_i350:
2132                 /* RX entries */
2133                 for (int i = 0; i < sc->rx_num_queues; i++) {
2134                         u32 index = i >> 1;
2135                         ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2136                         rx_que = &sc->rx_queues[i];
2137                         if (i & 1) {
2138                                 ivar &= 0xFF00FFFF;
2139                                 ivar |= (rx_que->msix | E1000_IVAR_VALID) << 16;
2140                         } else {
2141                                 ivar &= 0xFFFFFF00;
2142                                 ivar |= rx_que->msix | E1000_IVAR_VALID;
2143                         }
2144                         E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2145                 }
2146                 /* TX entries */
2147                 for (int i = 0; i < sc->tx_num_queues; i++) {
2148                         u32 index = i >> 1;
2149                         ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2150                         tx_que = &sc->tx_queues[i];
2151                         if (i & 1) {
2152                                 ivar &= 0x00FFFFFF;
2153                                 ivar |= (tx_que->msix | E1000_IVAR_VALID) << 24;
2154                         } else {
2155                                 ivar &= 0xFFFF00FF;
2156                                 ivar |= (tx_que->msix | E1000_IVAR_VALID) << 8;
2157                         }
2158                         E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2159                         sc->que_mask |= tx_que->eims;
2160                 }
2161
2162                 /* And for the link interrupt */
2163                 ivar = (sc->linkvec | E1000_IVAR_VALID) << 8;
2164                 sc->link_mask = 1 << sc->linkvec;
2165                 E1000_WRITE_REG(hw, E1000_IVAR_MISC, ivar);
2166                 break;
2167         case e1000_82576:
2168                 /* RX entries */
2169                 for (int i = 0; i < sc->rx_num_queues; i++) {
2170                         u32 index = i & 0x7; /* Each IVAR has two entries */
2171                         ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2172                         rx_que = &sc->rx_queues[i];
2173                         if (i < 8) {
2174                                 ivar &= 0xFFFFFF00;
2175                                 ivar |= rx_que->msix | E1000_IVAR_VALID;
2176                         } else {
2177                                 ivar &= 0xFF00FFFF;
2178                                 ivar |= (rx_que->msix | E1000_IVAR_VALID) << 16;
2179                         }
2180                         E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2181                         sc->que_mask |= rx_que->eims;
2182                 }
2183                 /* TX entries */
2184                 for (int i = 0; i < sc->tx_num_queues; i++) {
2185                         u32 index = i & 0x7; /* Each IVAR has two entries */
2186                         ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2187                         tx_que = &sc->tx_queues[i];
2188                         if (i < 8) {
2189                                 ivar &= 0xFFFF00FF;
2190                                 ivar |= (tx_que->msix | E1000_IVAR_VALID) << 8;
2191                         } else {
2192                                 ivar &= 0x00FFFFFF;
2193                                 ivar |= (tx_que->msix | E1000_IVAR_VALID) << 24;
2194                         }
2195                         E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2196                         sc->que_mask |= tx_que->eims;
2197                 }
2198
2199                 /* And for the link interrupt */
2200                 ivar = (sc->linkvec | E1000_IVAR_VALID) << 8;
2201                 sc->link_mask = 1 << sc->linkvec;
2202                 E1000_WRITE_REG(hw, E1000_IVAR_MISC, ivar);
2203                 break;
2204
2205         case e1000_82575:
2206                 /* enable MSI-X support*/
2207                 tmp = E1000_READ_REG(hw, E1000_CTRL_EXT);
2208                 tmp |= E1000_CTRL_EXT_PBA_CLR;
2209                 /* Auto-Mask interrupts upon ICR read. */
2210                 tmp |= E1000_CTRL_EXT_EIAME;
2211                 tmp |= E1000_CTRL_EXT_IRCA;
2212                 E1000_WRITE_REG(hw, E1000_CTRL_EXT, tmp);
2213
2214                 /* Queues */
2215                 for (int i = 0; i < sc->rx_num_queues; i++) {
2216                         rx_que = &sc->rx_queues[i];
2217                         tmp = E1000_EICR_RX_QUEUE0 << i;
2218                         tmp |= E1000_EICR_TX_QUEUE0 << i;
2219                         rx_que->eims = tmp;
2220                         E1000_WRITE_REG_ARRAY(hw, E1000_MSIXBM(0),
2221                             i, rx_que->eims);
2222                         sc->que_mask |= rx_que->eims;
2223                 }
2224
2225                 /* Link */
2226                 E1000_WRITE_REG(hw, E1000_MSIXBM(sc->linkvec),
2227                     E1000_EIMS_OTHER);
2228                 sc->link_mask |= E1000_EIMS_OTHER;
2229         default:
2230                 break;
2231         }
2232
2233         /* Set the starting interrupt rate */
2234         if (em_max_interrupt_rate > 0)
2235                 newitr = (4000000 / em_max_interrupt_rate) & 0x7FFC;
2236
2237         if (hw->mac.type == e1000_82575)
2238                 newitr |= newitr << 16;
2239         else
2240                 newitr |= E1000_EITR_CNT_IGNR;
2241
2242         for (int i = 0; i < sc->rx_num_queues; i++) {
2243                 rx_que = &sc->rx_queues[i];
2244                 E1000_WRITE_REG(hw, E1000_EITR(rx_que->msix), newitr);
2245         }
2246
2247         return;
2248 }
2249
2250 static void
2251 em_free_pci_resources(if_ctx_t ctx)
2252 {
2253         struct e1000_softc *sc = iflib_get_softc(ctx);
2254         struct em_rx_queue *que = sc->rx_queues;
2255         device_t dev = iflib_get_dev(ctx);
2256
2257         /* Release all MSI-X queue resources */
2258         if (sc->intr_type == IFLIB_INTR_MSIX)
2259                 iflib_irq_free(ctx, &sc->irq);
2260
2261         if (que != NULL) {
2262                 for (int i = 0; i < sc->rx_num_queues; i++, que++) {
2263                         iflib_irq_free(ctx, &que->que_irq);
2264                 }
2265         }
2266
2267         if (sc->memory != NULL) {
2268                 bus_release_resource(dev, SYS_RES_MEMORY,
2269                     rman_get_rid(sc->memory), sc->memory);
2270                 sc->memory = NULL;
2271         }
2272
2273         if (sc->flash != NULL) {
2274                 bus_release_resource(dev, SYS_RES_MEMORY,
2275                     rman_get_rid(sc->flash), sc->flash);
2276                 sc->flash = NULL;
2277         }
2278
2279         if (sc->ioport != NULL) {
2280                 bus_release_resource(dev, SYS_RES_IOPORT,
2281                     rman_get_rid(sc->ioport), sc->ioport);
2282                 sc->ioport = NULL;
2283         }
2284 }
2285
2286 /* Set up MSI or MSI-X */
2287 static int
2288 em_setup_msix(if_ctx_t ctx)
2289 {
2290         struct e1000_softc *sc = iflib_get_softc(ctx);
2291
2292         if (sc->hw.mac.type == e1000_82574) {
2293                 em_enable_vectors_82574(ctx);
2294         }
2295         return (0);
2296 }
2297
2298 /*********************************************************************
2299  *
2300  *  Workaround for SmartSpeed on 82541 and 82547 controllers
2301  *
2302  **********************************************************************/
2303 static void
2304 lem_smartspeed(struct e1000_softc *sc)
2305 {
2306         u16 phy_tmp;
2307
2308         if (sc->link_active || (sc->hw.phy.type != e1000_phy_igp) ||
2309             sc->hw.mac.autoneg == 0 ||
2310             (sc->hw.phy.autoneg_advertised & ADVERTISE_1000_FULL) == 0)
2311                 return;
2312
2313         if (sc->smartspeed == 0) {
2314                 /* If Master/Slave config fault is asserted twice,
2315                  * we assume back-to-back */
2316                 e1000_read_phy_reg(&sc->hw, PHY_1000T_STATUS, &phy_tmp);
2317                 if (!(phy_tmp & SR_1000T_MS_CONFIG_FAULT))
2318                         return;
2319                 e1000_read_phy_reg(&sc->hw, PHY_1000T_STATUS, &phy_tmp);
2320                 if (phy_tmp & SR_1000T_MS_CONFIG_FAULT) {
2321                         e1000_read_phy_reg(&sc->hw,
2322                             PHY_1000T_CTRL, &phy_tmp);
2323                         if(phy_tmp & CR_1000T_MS_ENABLE) {
2324                                 phy_tmp &= ~CR_1000T_MS_ENABLE;
2325                                 e1000_write_phy_reg(&sc->hw,
2326                                     PHY_1000T_CTRL, phy_tmp);
2327                                 sc->smartspeed++;
2328                                 if(sc->hw.mac.autoneg &&
2329                                    !e1000_copper_link_autoneg(&sc->hw) &&
2330                                    !e1000_read_phy_reg(&sc->hw,
2331                                     PHY_CONTROL, &phy_tmp)) {
2332                                         phy_tmp |= (MII_CR_AUTO_NEG_EN |
2333                                                     MII_CR_RESTART_AUTO_NEG);
2334                                         e1000_write_phy_reg(&sc->hw,
2335                                             PHY_CONTROL, phy_tmp);
2336                                 }
2337                         }
2338                 }
2339                 return;
2340         } else if(sc->smartspeed == EM_SMARTSPEED_DOWNSHIFT) {
2341                 /* If still no link, perhaps using 2/3 pair cable */
2342                 e1000_read_phy_reg(&sc->hw, PHY_1000T_CTRL, &phy_tmp);
2343                 phy_tmp |= CR_1000T_MS_ENABLE;
2344                 e1000_write_phy_reg(&sc->hw, PHY_1000T_CTRL, phy_tmp);
2345                 if(sc->hw.mac.autoneg &&
2346                    !e1000_copper_link_autoneg(&sc->hw) &&
2347                    !e1000_read_phy_reg(&sc->hw, PHY_CONTROL, &phy_tmp)) {
2348                         phy_tmp |= (MII_CR_AUTO_NEG_EN |
2349                                     MII_CR_RESTART_AUTO_NEG);
2350                         e1000_write_phy_reg(&sc->hw, PHY_CONTROL, phy_tmp);
2351                 }
2352         }
2353         /* Restart process after EM_SMARTSPEED_MAX iterations */
2354         if(sc->smartspeed++ == EM_SMARTSPEED_MAX)
2355                 sc->smartspeed = 0;
2356 }
2357
2358 /*********************************************************************
2359  *
2360  *  Initialize the DMA Coalescing feature
2361  *
2362  **********************************************************************/
2363 static void
2364 igb_init_dmac(struct e1000_softc *sc, u32 pba)
2365 {
2366         device_t        dev = sc->dev;
2367         struct e1000_hw *hw = &sc->hw;
2368         u32             dmac, reg = ~E1000_DMACR_DMAC_EN;
2369         u16             hwm;
2370         u16             max_frame_size;
2371
2372         if (hw->mac.type == e1000_i211)
2373                 return;
2374
2375         max_frame_size = sc->shared->isc_max_frame_size;
2376         if (hw->mac.type > e1000_82580) {
2377
2378                 if (sc->dmac == 0) { /* Disabling it */
2379                         E1000_WRITE_REG(hw, E1000_DMACR, reg);
2380                         return;
2381                 } else
2382                         device_printf(dev, "DMA Coalescing enabled\n");
2383
2384                 /* Set starting threshold */
2385                 E1000_WRITE_REG(hw, E1000_DMCTXTH, 0);
2386
2387                 hwm = 64 * pba - max_frame_size / 16;
2388                 if (hwm < 64 * (pba - 6))
2389                         hwm = 64 * (pba - 6);
2390                 reg = E1000_READ_REG(hw, E1000_FCRTC);
2391                 reg &= ~E1000_FCRTC_RTH_COAL_MASK;
2392                 reg |= ((hwm << E1000_FCRTC_RTH_COAL_SHIFT)
2393                     & E1000_FCRTC_RTH_COAL_MASK);
2394                 E1000_WRITE_REG(hw, E1000_FCRTC, reg);
2395
2396
2397                 dmac = pba - max_frame_size / 512;
2398                 if (dmac < pba - 10)
2399                         dmac = pba - 10;
2400                 reg = E1000_READ_REG(hw, E1000_DMACR);
2401                 reg &= ~E1000_DMACR_DMACTHR_MASK;
2402                 reg |= ((dmac << E1000_DMACR_DMACTHR_SHIFT)
2403                     & E1000_DMACR_DMACTHR_MASK);
2404
2405                 /* transition to L0x or L1 if available..*/
2406                 reg |= (E1000_DMACR_DMAC_EN | E1000_DMACR_DMAC_LX_MASK);
2407
2408                 /* Check if status is 2.5Gb backplane connection
2409                 * before configuration of watchdog timer, which is
2410                 * in msec values in 12.8usec intervals
2411                 * watchdog timer= msec values in 32usec intervals
2412                 * for non 2.5Gb connection
2413                 */
2414                 if (hw->mac.type == e1000_i354) {
2415                         int status = E1000_READ_REG(hw, E1000_STATUS);
2416                         if ((status & E1000_STATUS_2P5_SKU) &&
2417                             (!(status & E1000_STATUS_2P5_SKU_OVER)))
2418                                 reg |= ((sc->dmac * 5) >> 6);
2419                         else
2420                                 reg |= (sc->dmac >> 5);
2421                 } else {
2422                         reg |= (sc->dmac >> 5);
2423                 }
2424
2425                 E1000_WRITE_REG(hw, E1000_DMACR, reg);
2426
2427                 E1000_WRITE_REG(hw, E1000_DMCRTRH, 0);
2428
2429                 /* Set the interval before transition */
2430                 reg = E1000_READ_REG(hw, E1000_DMCTLX);
2431                 if (hw->mac.type == e1000_i350)
2432                         reg |= IGB_DMCTLX_DCFLUSH_DIS;
2433                 /*
2434                 ** in 2.5Gb connection, TTLX unit is 0.4 usec
2435                 ** which is 0x4*2 = 0xA. But delay is still 4 usec
2436                 */
2437                 if (hw->mac.type == e1000_i354) {
2438                         int status = E1000_READ_REG(hw, E1000_STATUS);
2439                         if ((status & E1000_STATUS_2P5_SKU) &&
2440                             (!(status & E1000_STATUS_2P5_SKU_OVER)))
2441                                 reg |= 0xA;
2442                         else
2443                                 reg |= 0x4;
2444                 } else {
2445                         reg |= 0x4;
2446                 }
2447
2448                 E1000_WRITE_REG(hw, E1000_DMCTLX, reg);
2449
2450                 /* free space in tx packet buffer to wake from DMA coal */
2451                 E1000_WRITE_REG(hw, E1000_DMCTXTH, (IGB_TXPBSIZE -
2452                     (2 * max_frame_size)) >> 6);
2453
2454                 /* make low power state decision controlled by DMA coal */
2455                 reg = E1000_READ_REG(hw, E1000_PCIEMISC);
2456                 reg &= ~E1000_PCIEMISC_LX_DECISION;
2457                 E1000_WRITE_REG(hw, E1000_PCIEMISC, reg);
2458
2459         } else if (hw->mac.type == e1000_82580) {
2460                 u32 reg = E1000_READ_REG(hw, E1000_PCIEMISC);
2461                 E1000_WRITE_REG(hw, E1000_PCIEMISC,
2462                     reg & ~E1000_PCIEMISC_LX_DECISION);
2463                 E1000_WRITE_REG(hw, E1000_DMACR, 0);
2464         }
2465 }
2466
2467 /*********************************************************************
2468  *
2469  *  Initialize the hardware to a configuration as specified by the
2470  *  sc structure.
2471  *
2472  **********************************************************************/
2473 static void
2474 em_reset(if_ctx_t ctx)
2475 {
2476         device_t dev = iflib_get_dev(ctx);
2477         struct e1000_softc *sc = iflib_get_softc(ctx);
2478         struct ifnet *ifp = iflib_get_ifp(ctx);
2479         struct e1000_hw *hw = &sc->hw;
2480         u32 rx_buffer_size;
2481         u32 pba;
2482
2483         INIT_DEBUGOUT("em_reset: begin");
2484         /* Let the firmware know the OS is in control */
2485         em_get_hw_control(sc);
2486
2487         /* Set up smart power down as default off on newer adapters. */
2488         if (!em_smart_pwr_down && (hw->mac.type == e1000_82571 ||
2489             hw->mac.type == e1000_82572)) {
2490                 u16 phy_tmp = 0;
2491
2492                 /* Speed up time to link by disabling smart power down. */
2493                 e1000_read_phy_reg(hw, IGP02E1000_PHY_POWER_MGMT, &phy_tmp);
2494                 phy_tmp &= ~IGP02E1000_PM_SPD;
2495                 e1000_write_phy_reg(hw, IGP02E1000_PHY_POWER_MGMT, phy_tmp);
2496         }
2497
2498         /*
2499          * Packet Buffer Allocation (PBA)
2500          * Writing PBA sets the receive portion of the buffer
2501          * the remainder is used for the transmit buffer.
2502          */
2503         switch (hw->mac.type) {
2504         /* 82547: Total Packet Buffer is 40K */
2505         case e1000_82547:
2506         case e1000_82547_rev_2:
2507                 if (hw->mac.max_frame_size > 8192)
2508                         pba = E1000_PBA_22K; /* 22K for Rx, 18K for Tx */
2509                 else
2510                         pba = E1000_PBA_30K; /* 30K for Rx, 10K for Tx */
2511                 break;
2512         /* 82571/82572/80003es2lan: Total Packet Buffer is 48K */
2513         case e1000_82571:
2514         case e1000_82572:
2515         case e1000_80003es2lan:
2516                         pba = E1000_PBA_32K; /* 32K for Rx, 16K for Tx */
2517                 break;
2518         /* 82573: Total Packet Buffer is 32K */
2519         case e1000_82573:
2520                         pba = E1000_PBA_12K; /* 12K for Rx, 20K for Tx */
2521                 break;
2522         case e1000_82574:
2523         case e1000_82583:
2524                         pba = E1000_PBA_20K; /* 20K for Rx, 20K for Tx */
2525                 break;
2526         case e1000_ich8lan:
2527                 pba = E1000_PBA_8K;
2528                 break;
2529         case e1000_ich9lan:
2530         case e1000_ich10lan:
2531                 /* Boost Receive side for jumbo frames */
2532                 if (hw->mac.max_frame_size > 4096)
2533                         pba = E1000_PBA_14K;
2534                 else
2535                         pba = E1000_PBA_10K;
2536                 break;
2537         case e1000_pchlan:
2538         case e1000_pch2lan:
2539         case e1000_pch_lpt:
2540         case e1000_pch_spt:
2541         case e1000_pch_cnp:
2542         case e1000_pch_tgp:
2543         case e1000_pch_adp:
2544         case e1000_pch_mtp:
2545                 pba = E1000_PBA_26K;
2546                 break;
2547         case e1000_82575:
2548                 pba = E1000_PBA_32K;
2549                 break;
2550         case e1000_82576:
2551         case e1000_vfadapt:
2552                 pba = E1000_READ_REG(hw, E1000_RXPBS);
2553                 pba &= E1000_RXPBS_SIZE_MASK_82576;
2554                 break;
2555         case e1000_82580:
2556         case e1000_i350:
2557         case e1000_i354:
2558         case e1000_vfadapt_i350:
2559                 pba = E1000_READ_REG(hw, E1000_RXPBS);
2560                 pba = e1000_rxpbs_adjust_82580(pba);
2561                 break;
2562         case e1000_i210:
2563         case e1000_i211:
2564                 pba = E1000_PBA_34K;
2565                 break;
2566         default:
2567                 /* Remaining devices assumed to have a Packet Buffer of 64K. */
2568                 if (hw->mac.max_frame_size > 8192)
2569                         pba = E1000_PBA_40K; /* 40K for Rx, 24K for Tx */
2570                 else
2571                         pba = E1000_PBA_48K; /* 48K for Rx, 16K for Tx */
2572         }
2573
2574         /* Special needs in case of Jumbo frames */
2575         if ((hw->mac.type == e1000_82575) && (ifp->if_mtu > ETHERMTU)) {
2576                 u32 tx_space, min_tx, min_rx;
2577                 pba = E1000_READ_REG(hw, E1000_PBA);
2578                 tx_space = pba >> 16;
2579                 pba &= 0xffff;
2580                 min_tx = (hw->mac.max_frame_size +
2581                     sizeof(struct e1000_tx_desc) - ETHERNET_FCS_SIZE) * 2;
2582                 min_tx = roundup2(min_tx, 1024);
2583                 min_tx >>= 10;
2584                 min_rx = hw->mac.max_frame_size;
2585                 min_rx = roundup2(min_rx, 1024);
2586                 min_rx >>= 10;
2587                 if (tx_space < min_tx &&
2588                     ((min_tx - tx_space) < pba)) {
2589                         pba = pba - (min_tx - tx_space);
2590                         /*
2591                          * if short on rx space, rx wins
2592                          * and must trump tx adjustment
2593                          */
2594                         if (pba < min_rx)
2595                                 pba = min_rx;
2596                 }
2597                 E1000_WRITE_REG(hw, E1000_PBA, pba);
2598         }
2599
2600         if (hw->mac.type < igb_mac_min)
2601                 E1000_WRITE_REG(hw, E1000_PBA, pba);
2602
2603         INIT_DEBUGOUT1("em_reset: pba=%dK",pba);
2604
2605         /*
2606          * These parameters control the automatic generation (Tx) and
2607          * response (Rx) to Ethernet PAUSE frames.
2608          * - High water mark should allow for at least two frames to be
2609          *   received after sending an XOFF.
2610          * - Low water mark works best when it is very near the high water mark.
2611          *   This allows the receiver to restart by sending XON when it has
2612          *   drained a bit. Here we use an arbitrary value of 1500 which will
2613          *   restart after one full frame is pulled from the buffer. There
2614          *   could be several smaller frames in the buffer and if so they will
2615          *   not trigger the XON until their total number reduces the buffer
2616          *   by 1500.
2617          * - The pause time is fairly large at 1000 x 512ns = 512 usec.
2618          */
2619         rx_buffer_size = (pba & 0xffff) << 10;
2620         hw->fc.high_water = rx_buffer_size -
2621             roundup2(hw->mac.max_frame_size, 1024);
2622         hw->fc.low_water = hw->fc.high_water - 1500;
2623
2624         if (sc->fc) /* locally set flow control value? */
2625                 hw->fc.requested_mode = sc->fc;
2626         else
2627                 hw->fc.requested_mode = e1000_fc_full;
2628
2629         if (hw->mac.type == e1000_80003es2lan)
2630                 hw->fc.pause_time = 0xFFFF;
2631         else
2632                 hw->fc.pause_time = EM_FC_PAUSE_TIME;
2633
2634         hw->fc.send_xon = true;
2635
2636         /* Device specific overrides/settings */
2637         switch (hw->mac.type) {
2638         case e1000_pchlan:
2639                 /* Workaround: no TX flow ctrl for PCH */
2640                 hw->fc.requested_mode = e1000_fc_rx_pause;
2641                 hw->fc.pause_time = 0xFFFF; /* override */
2642                 if (if_getmtu(ifp) > ETHERMTU) {
2643                         hw->fc.high_water = 0x3500;
2644                         hw->fc.low_water = 0x1500;
2645                 } else {
2646                         hw->fc.high_water = 0x5000;
2647                         hw->fc.low_water = 0x3000;
2648                 }
2649                 hw->fc.refresh_time = 0x1000;
2650                 break;
2651         case e1000_pch2lan:
2652         case e1000_pch_lpt:
2653         case e1000_pch_spt:
2654         case e1000_pch_cnp:
2655         case e1000_pch_tgp:
2656         case e1000_pch_adp:
2657         case e1000_pch_mtp:
2658                 hw->fc.high_water = 0x5C20;
2659                 hw->fc.low_water = 0x5048;
2660                 hw->fc.pause_time = 0x0650;
2661                 hw->fc.refresh_time = 0x0400;
2662                 /* Jumbos need adjusted PBA */
2663                 if (if_getmtu(ifp) > ETHERMTU)
2664                         E1000_WRITE_REG(hw, E1000_PBA, 12);
2665                 else
2666                         E1000_WRITE_REG(hw, E1000_PBA, 26);
2667                 break;
2668         case e1000_82575:
2669         case e1000_82576:
2670                 /* 8-byte granularity */
2671                 hw->fc.low_water = hw->fc.high_water - 8;
2672                 break;
2673         case e1000_82580:
2674         case e1000_i350:
2675         case e1000_i354:
2676         case e1000_i210:
2677         case e1000_i211:
2678         case e1000_vfadapt:
2679         case e1000_vfadapt_i350:
2680                 /* 16-byte granularity */
2681                 hw->fc.low_water = hw->fc.high_water - 16;
2682                 break;
2683         case e1000_ich9lan:
2684         case e1000_ich10lan:
2685                 if (if_getmtu(ifp) > ETHERMTU) {
2686                         hw->fc.high_water = 0x2800;
2687                         hw->fc.low_water = hw->fc.high_water - 8;
2688                         break;
2689                 }
2690                 /* FALLTHROUGH */
2691         default:
2692                 if (hw->mac.type == e1000_80003es2lan)
2693                         hw->fc.pause_time = 0xFFFF;
2694                 break;
2695         }
2696
2697         /* Issue a global reset */
2698         e1000_reset_hw(hw);
2699         if (hw->mac.type >= igb_mac_min) {
2700                 E1000_WRITE_REG(hw, E1000_WUC, 0);
2701         } else {
2702                 E1000_WRITE_REG(hw, E1000_WUFC, 0);
2703                 em_disable_aspm(sc);
2704         }
2705         if (sc->flags & IGB_MEDIA_RESET) {
2706                 e1000_setup_init_funcs(hw, true);
2707                 e1000_get_bus_info(hw);
2708                 sc->flags &= ~IGB_MEDIA_RESET;
2709         }
2710         /* and a re-init */
2711         if (e1000_init_hw(hw) < 0) {
2712                 device_printf(dev, "Hardware Initialization Failed\n");
2713                 return;
2714         }
2715         if (hw->mac.type >= igb_mac_min)
2716                 igb_init_dmac(sc, pba);
2717
2718         E1000_WRITE_REG(hw, E1000_VET, ETHERTYPE_VLAN);
2719         e1000_get_phy_info(hw);
2720         e1000_check_for_link(hw);
2721 }
2722
2723 /*
2724  * Initialise the RSS mapping for NICs that support multiple transmit/
2725  * receive rings.
2726  */
2727
2728 #define RSSKEYLEN 10
2729 static void
2730 em_initialize_rss_mapping(struct e1000_softc *sc)
2731 {
2732         uint8_t  rss_key[4 * RSSKEYLEN];
2733         uint32_t reta = 0;
2734         struct e1000_hw *hw = &sc->hw;
2735         int i;
2736
2737         /*
2738          * Configure RSS key
2739          */
2740         arc4rand(rss_key, sizeof(rss_key), 0);
2741         for (i = 0; i < RSSKEYLEN; ++i) {
2742                 uint32_t rssrk = 0;
2743
2744                 rssrk = EM_RSSRK_VAL(rss_key, i);
2745                 E1000_WRITE_REG(hw,E1000_RSSRK(i), rssrk);
2746         }
2747
2748         /*
2749          * Configure RSS redirect table in following fashion:
2750          * (hash & ring_cnt_mask) == rdr_table[(hash & rdr_table_mask)]
2751          */
2752         for (i = 0; i < sizeof(reta); ++i) {
2753                 uint32_t q;
2754
2755                 q = (i % sc->rx_num_queues) << 7;
2756                 reta |= q << (8 * i);
2757         }
2758
2759         for (i = 0; i < 32; ++i)
2760                 E1000_WRITE_REG(hw, E1000_RETA(i), reta);
2761
2762         E1000_WRITE_REG(hw, E1000_MRQC, E1000_MRQC_RSS_ENABLE_2Q |
2763                         E1000_MRQC_RSS_FIELD_IPV4_TCP |
2764                         E1000_MRQC_RSS_FIELD_IPV4 |
2765                         E1000_MRQC_RSS_FIELD_IPV6_TCP_EX |
2766                         E1000_MRQC_RSS_FIELD_IPV6_EX |
2767                         E1000_MRQC_RSS_FIELD_IPV6);
2768 }
2769
2770 static void
2771 igb_initialize_rss_mapping(struct e1000_softc *sc)
2772 {
2773         struct e1000_hw *hw = &sc->hw;
2774         int i;
2775         int queue_id;
2776         u32 reta;
2777         u32 rss_key[10], mrqc, shift = 0;
2778
2779         /* XXX? */
2780         if (hw->mac.type == e1000_82575)
2781                 shift = 6;
2782
2783         /*
2784          * The redirection table controls which destination
2785          * queue each bucket redirects traffic to.
2786          * Each DWORD represents four queues, with the LSB
2787          * being the first queue in the DWORD.
2788          *
2789          * This just allocates buckets to queues using round-robin
2790          * allocation.
2791          *
2792          * NOTE: It Just Happens to line up with the default
2793          * RSS allocation method.
2794          */
2795
2796         /* Warning FM follows */
2797         reta = 0;
2798         for (i = 0; i < 128; i++) {
2799 #ifdef RSS
2800                 queue_id = rss_get_indirection_to_bucket(i);
2801                 /*
2802                  * If we have more queues than buckets, we'll
2803                  * end up mapping buckets to a subset of the
2804                  * queues.
2805                  *
2806                  * If we have more buckets than queues, we'll
2807                  * end up instead assigning multiple buckets
2808                  * to queues.
2809                  *
2810                  * Both are suboptimal, but we need to handle
2811                  * the case so we don't go out of bounds
2812                  * indexing arrays and such.
2813                  */
2814                 queue_id = queue_id % sc->rx_num_queues;
2815 #else
2816                 queue_id = (i % sc->rx_num_queues);
2817 #endif
2818                 /* Adjust if required */
2819                 queue_id = queue_id << shift;
2820
2821                 /*
2822                  * The low 8 bits are for hash value (n+0);
2823                  * The next 8 bits are for hash value (n+1), etc.
2824                  */
2825                 reta = reta >> 8;
2826                 reta = reta | ( ((uint32_t) queue_id) << 24);
2827                 if ((i & 3) == 3) {
2828                         E1000_WRITE_REG(hw, E1000_RETA(i >> 2), reta);
2829                         reta = 0;
2830                 }
2831         }
2832
2833         /* Now fill in hash table */
2834
2835         /*
2836          * MRQC: Multiple Receive Queues Command
2837          * Set queuing to RSS control, number depends on the device.
2838          */
2839         mrqc = E1000_MRQC_ENABLE_RSS_MQ;
2840
2841 #ifdef RSS
2842         /* XXX ew typecasting */
2843         rss_getkey((uint8_t *) &rss_key);
2844 #else
2845         arc4rand(&rss_key, sizeof(rss_key), 0);
2846 #endif
2847         for (i = 0; i < 10; i++)
2848                 E1000_WRITE_REG_ARRAY(hw, E1000_RSSRK(0), i, rss_key[i]);
2849
2850         /*
2851          * Configure the RSS fields to hash upon.
2852          */
2853         mrqc |= (E1000_MRQC_RSS_FIELD_IPV4 |
2854             E1000_MRQC_RSS_FIELD_IPV4_TCP);
2855         mrqc |= (E1000_MRQC_RSS_FIELD_IPV6 |
2856             E1000_MRQC_RSS_FIELD_IPV6_TCP);
2857         mrqc |=( E1000_MRQC_RSS_FIELD_IPV4_UDP |
2858             E1000_MRQC_RSS_FIELD_IPV6_UDP);
2859         mrqc |=( E1000_MRQC_RSS_FIELD_IPV6_UDP_EX |
2860             E1000_MRQC_RSS_FIELD_IPV6_TCP_EX);
2861
2862         E1000_WRITE_REG(hw, E1000_MRQC, mrqc);
2863 }
2864
2865 /*********************************************************************
2866  *
2867  *  Setup networking device structure and register interface media.
2868  *
2869  **********************************************************************/
2870 static int
2871 em_setup_interface(if_ctx_t ctx)
2872 {
2873         struct ifnet *ifp = iflib_get_ifp(ctx);
2874         struct e1000_softc *sc = iflib_get_softc(ctx);
2875         if_softc_ctx_t scctx = sc->shared;
2876
2877         INIT_DEBUGOUT("em_setup_interface: begin");
2878
2879         /* Single Queue */
2880         if (sc->tx_num_queues == 1) {
2881                 if_setsendqlen(ifp, scctx->isc_ntxd[0] - 1);
2882                 if_setsendqready(ifp);
2883         }
2884
2885         /*
2886          * Specify the media types supported by this adapter and register
2887          * callbacks to update media and link information
2888          */
2889         if (sc->hw.phy.media_type == e1000_media_type_fiber ||
2890             sc->hw.phy.media_type == e1000_media_type_internal_serdes) {
2891                 u_char fiber_type = IFM_1000_SX;        /* default type */
2892
2893                 if (sc->hw.mac.type == e1000_82545)
2894                         fiber_type = IFM_1000_LX;
2895                 ifmedia_add(sc->media, IFM_ETHER | fiber_type | IFM_FDX, 0, NULL);
2896                 ifmedia_add(sc->media, IFM_ETHER | fiber_type, 0, NULL);
2897         } else {
2898                 ifmedia_add(sc->media, IFM_ETHER | IFM_10_T, 0, NULL);
2899                 ifmedia_add(sc->media, IFM_ETHER | IFM_10_T | IFM_FDX, 0, NULL);
2900                 ifmedia_add(sc->media, IFM_ETHER | IFM_100_TX, 0, NULL);
2901                 ifmedia_add(sc->media, IFM_ETHER | IFM_100_TX | IFM_FDX, 0, NULL);
2902                 if (sc->hw.phy.type != e1000_phy_ife) {
2903                         ifmedia_add(sc->media, IFM_ETHER | IFM_1000_T | IFM_FDX, 0, NULL);
2904                         ifmedia_add(sc->media, IFM_ETHER | IFM_1000_T, 0, NULL);
2905                 }
2906         }
2907         ifmedia_add(sc->media, IFM_ETHER | IFM_AUTO, 0, NULL);
2908         ifmedia_set(sc->media, IFM_ETHER | IFM_AUTO);
2909         return (0);
2910 }
2911
2912 static int
2913 em_if_tx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int ntxqs, int ntxqsets)
2914 {
2915         struct e1000_softc *sc = iflib_get_softc(ctx);
2916         if_softc_ctx_t scctx = sc->shared;
2917         int error = E1000_SUCCESS;
2918         struct em_tx_queue *que;
2919         int i, j;
2920
2921         MPASS(sc->tx_num_queues > 0);
2922         MPASS(sc->tx_num_queues == ntxqsets);
2923
2924         /* First allocate the top level queue structs */
2925         if (!(sc->tx_queues =
2926             (struct em_tx_queue *) malloc(sizeof(struct em_tx_queue) *
2927             sc->tx_num_queues, M_DEVBUF, M_NOWAIT | M_ZERO))) {
2928                 device_printf(iflib_get_dev(ctx), "Unable to allocate queue memory\n");
2929                 return(ENOMEM);
2930         }
2931
2932         for (i = 0, que = sc->tx_queues; i < sc->tx_num_queues; i++, que++) {
2933                 /* Set up some basics */
2934
2935                 struct tx_ring *txr = &que->txr;
2936                 txr->sc = que->sc = sc;
2937                 que->me = txr->me =  i;
2938
2939                 /* Allocate report status array */
2940                 if (!(txr->tx_rsq = (qidx_t *) malloc(sizeof(qidx_t) * scctx->isc_ntxd[0], M_DEVBUF, M_NOWAIT | M_ZERO))) {
2941                         device_printf(iflib_get_dev(ctx), "failed to allocate rs_idxs memory\n");
2942                         error = ENOMEM;
2943                         goto fail;
2944                 }
2945                 for (j = 0; j < scctx->isc_ntxd[0]; j++)
2946                         txr->tx_rsq[j] = QIDX_INVALID;
2947                 /* get the virtual and physical address of the hardware queues */
2948                 txr->tx_base = (struct e1000_tx_desc *)vaddrs[i*ntxqs];
2949                 txr->tx_paddr = paddrs[i*ntxqs];
2950         }
2951
2952         if (bootverbose)
2953                 device_printf(iflib_get_dev(ctx),
2954                     "allocated for %d tx_queues\n", sc->tx_num_queues);
2955         return (0);
2956 fail:
2957         em_if_queues_free(ctx);
2958         return (error);
2959 }
2960
2961 static int
2962 em_if_rx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int nrxqs, int nrxqsets)
2963 {
2964         struct e1000_softc *sc = iflib_get_softc(ctx);
2965         int error = E1000_SUCCESS;
2966         struct em_rx_queue *que;
2967         int i;
2968
2969         MPASS(sc->rx_num_queues > 0);
2970         MPASS(sc->rx_num_queues == nrxqsets);
2971
2972         /* First allocate the top level queue structs */
2973         if (!(sc->rx_queues =
2974             (struct em_rx_queue *) malloc(sizeof(struct em_rx_queue) *
2975             sc->rx_num_queues, M_DEVBUF, M_NOWAIT | M_ZERO))) {
2976                 device_printf(iflib_get_dev(ctx), "Unable to allocate queue memory\n");
2977                 error = ENOMEM;
2978                 goto fail;
2979         }
2980
2981         for (i = 0, que = sc->rx_queues; i < nrxqsets; i++, que++) {
2982                 /* Set up some basics */
2983                 struct rx_ring *rxr = &que->rxr;
2984                 rxr->sc = que->sc = sc;
2985                 rxr->que = que;
2986                 que->me = rxr->me =  i;
2987
2988                 /* get the virtual and physical address of the hardware queues */
2989                 rxr->rx_base = (union e1000_rx_desc_extended *)vaddrs[i*nrxqs];
2990                 rxr->rx_paddr = paddrs[i*nrxqs];
2991         }
2992  
2993         if (bootverbose)
2994                 device_printf(iflib_get_dev(ctx),
2995                     "allocated for %d rx_queues\n", sc->rx_num_queues);
2996
2997         return (0);
2998 fail:
2999         em_if_queues_free(ctx);
3000         return (error);
3001 }
3002
3003 static void
3004 em_if_queues_free(if_ctx_t ctx)
3005 {
3006         struct e1000_softc *sc = iflib_get_softc(ctx);
3007         struct em_tx_queue *tx_que = sc->tx_queues;
3008         struct em_rx_queue *rx_que = sc->rx_queues;
3009
3010         if (tx_que != NULL) {
3011                 for (int i = 0; i < sc->tx_num_queues; i++, tx_que++) {
3012                         struct tx_ring *txr = &tx_que->txr;
3013                         if (txr->tx_rsq == NULL)
3014                                 break;
3015
3016                         free(txr->tx_rsq, M_DEVBUF);
3017                         txr->tx_rsq = NULL;
3018                 }
3019                 free(sc->tx_queues, M_DEVBUF);
3020                 sc->tx_queues = NULL;
3021         }
3022
3023         if (rx_que != NULL) {
3024                 free(sc->rx_queues, M_DEVBUF);
3025                 sc->rx_queues = NULL;
3026         }
3027 }
3028
3029 /*********************************************************************
3030  *
3031  *  Enable transmit unit.
3032  *
3033  **********************************************************************/
3034 static void
3035 em_initialize_transmit_unit(if_ctx_t ctx)
3036 {
3037         struct e1000_softc *sc = iflib_get_softc(ctx);
3038         if_softc_ctx_t scctx = sc->shared;
3039         struct em_tx_queue *que;
3040         struct tx_ring  *txr;
3041         struct e1000_hw *hw = &sc->hw;
3042         u32 tctl, txdctl = 0, tarc, tipg = 0;
3043
3044         INIT_DEBUGOUT("em_initialize_transmit_unit: begin");
3045
3046         for (int i = 0; i < sc->tx_num_queues; i++, txr++) {
3047                 u64 bus_addr;
3048                 caddr_t offp, endp;
3049
3050                 que = &sc->tx_queues[i];
3051                 txr = &que->txr;
3052                 bus_addr = txr->tx_paddr;
3053
3054                 /* Clear checksum offload context. */
3055                 offp = (caddr_t)&txr->csum_flags;
3056                 endp = (caddr_t)(txr + 1);
3057                 bzero(offp, endp - offp);
3058
3059                 /* Base and Len of TX Ring */
3060                 E1000_WRITE_REG(hw, E1000_TDLEN(i),
3061                     scctx->isc_ntxd[0] * sizeof(struct e1000_tx_desc));
3062                 E1000_WRITE_REG(hw, E1000_TDBAH(i),
3063                     (u32)(bus_addr >> 32));
3064                 E1000_WRITE_REG(hw, E1000_TDBAL(i),
3065                     (u32)bus_addr);
3066                 /* Init the HEAD/TAIL indices */
3067                 E1000_WRITE_REG(hw, E1000_TDT(i), 0);
3068                 E1000_WRITE_REG(hw, E1000_TDH(i), 0);
3069
3070                 HW_DEBUGOUT2("Base = %x, Length = %x\n",
3071                     E1000_READ_REG(hw, E1000_TDBAL(i)),
3072                     E1000_READ_REG(hw, E1000_TDLEN(i)));
3073
3074                 txdctl = 0; /* clear txdctl */
3075                 txdctl |= 0x1f; /* PTHRESH */
3076                 txdctl |= 1 << 8; /* HTHRESH */
3077                 txdctl |= 1 << 16;/* WTHRESH */
3078                 txdctl |= 1 << 22; /* Reserved bit 22 must always be 1 */
3079                 txdctl |= E1000_TXDCTL_GRAN;
3080                 txdctl |= 1 << 25; /* LWTHRESH */
3081
3082                 E1000_WRITE_REG(hw, E1000_TXDCTL(i), txdctl);
3083         }
3084
3085         /* Set the default values for the Tx Inter Packet Gap timer */
3086         switch (hw->mac.type) {
3087         case e1000_80003es2lan:
3088                 tipg = DEFAULT_82543_TIPG_IPGR1;
3089                 tipg |= DEFAULT_80003ES2LAN_TIPG_IPGR2 <<
3090                     E1000_TIPG_IPGR2_SHIFT;
3091                 break;
3092         case e1000_82542:
3093                 tipg = DEFAULT_82542_TIPG_IPGT;
3094                 tipg |= DEFAULT_82542_TIPG_IPGR1 << E1000_TIPG_IPGR1_SHIFT;
3095                 tipg |= DEFAULT_82542_TIPG_IPGR2 << E1000_TIPG_IPGR2_SHIFT;
3096                 break;
3097         default:
3098                 if (hw->phy.media_type == e1000_media_type_fiber ||
3099                     hw->phy.media_type == e1000_media_type_internal_serdes)
3100                         tipg = DEFAULT_82543_TIPG_IPGT_FIBER;
3101                 else
3102                         tipg = DEFAULT_82543_TIPG_IPGT_COPPER;
3103                 tipg |= DEFAULT_82543_TIPG_IPGR1 << E1000_TIPG_IPGR1_SHIFT;
3104                 tipg |= DEFAULT_82543_TIPG_IPGR2 << E1000_TIPG_IPGR2_SHIFT;
3105         }
3106
3107         E1000_WRITE_REG(hw, E1000_TIPG, tipg);
3108         E1000_WRITE_REG(hw, E1000_TIDV, sc->tx_int_delay.value);
3109
3110         if(hw->mac.type >= e1000_82540)
3111                 E1000_WRITE_REG(hw, E1000_TADV,
3112                     sc->tx_abs_int_delay.value);
3113
3114         if (hw->mac.type == e1000_82571 || hw->mac.type == e1000_82572) {
3115                 tarc = E1000_READ_REG(hw, E1000_TARC(0));
3116                 tarc |= TARC_SPEED_MODE_BIT;
3117                 E1000_WRITE_REG(hw, E1000_TARC(0), tarc);
3118         } else if (hw->mac.type == e1000_80003es2lan) {
3119                 /* errata: program both queues to unweighted RR */
3120                 tarc = E1000_READ_REG(hw, E1000_TARC(0));
3121                 tarc |= 1;
3122                 E1000_WRITE_REG(hw, E1000_TARC(0), tarc);
3123                 tarc = E1000_READ_REG(hw, E1000_TARC(1));
3124                 tarc |= 1;
3125                 E1000_WRITE_REG(hw, E1000_TARC(1), tarc);
3126         } else if (hw->mac.type == e1000_82574) {
3127                 tarc = E1000_READ_REG(hw, E1000_TARC(0));
3128                 tarc |= TARC_ERRATA_BIT;
3129                 if ( sc->tx_num_queues > 1) {
3130                         tarc |= (TARC_COMPENSATION_MODE | TARC_MQ_FIX);
3131                         E1000_WRITE_REG(hw, E1000_TARC(0), tarc);
3132                         E1000_WRITE_REG(hw, E1000_TARC(1), tarc);
3133                 } else
3134                         E1000_WRITE_REG(hw, E1000_TARC(0), tarc);
3135         }
3136
3137         if (sc->tx_int_delay.value > 0)
3138                 sc->txd_cmd |= E1000_TXD_CMD_IDE;
3139
3140         /* Program the Transmit Control Register */
3141         tctl = E1000_READ_REG(hw, E1000_TCTL);
3142         tctl &= ~E1000_TCTL_CT;
3143         tctl |= (E1000_TCTL_PSP | E1000_TCTL_RTLC | E1000_TCTL_EN |
3144                    (E1000_COLLISION_THRESHOLD << E1000_CT_SHIFT));
3145
3146         if (hw->mac.type >= e1000_82571)
3147                 tctl |= E1000_TCTL_MULR;
3148
3149         /* This write will effectively turn on the transmit unit. */
3150         E1000_WRITE_REG(hw, E1000_TCTL, tctl);
3151
3152         /* SPT and KBL errata workarounds */
3153         if (hw->mac.type == e1000_pch_spt) {
3154                 u32 reg;
3155                 reg = E1000_READ_REG(hw, E1000_IOSFPC);
3156                 reg |= E1000_RCTL_RDMTS_HEX;
3157                 E1000_WRITE_REG(hw, E1000_IOSFPC, reg);
3158                 /* i218-i219 Specification Update 1.5.4.5 */
3159                 reg = E1000_READ_REG(hw, E1000_TARC(0));
3160                 reg &= ~E1000_TARC0_CB_MULTIQ_3_REQ;
3161                 reg |= E1000_TARC0_CB_MULTIQ_2_REQ;
3162                 E1000_WRITE_REG(hw, E1000_TARC(0), reg);
3163         }
3164 }
3165
3166 /*********************************************************************
3167  *
3168  *  Enable receive unit.
3169  *
3170  **********************************************************************/
3171 #define BSIZEPKT_ROUNDUP ((1<<E1000_SRRCTL_BSIZEPKT_SHIFT)-1)
3172
3173 static void
3174 em_initialize_receive_unit(if_ctx_t ctx)
3175 {
3176         struct e1000_softc *sc = iflib_get_softc(ctx);
3177         if_softc_ctx_t scctx = sc->shared;
3178         struct ifnet *ifp = iflib_get_ifp(ctx);
3179         struct e1000_hw *hw = &sc->hw;
3180         struct em_rx_queue *que;
3181         int i;
3182         uint32_t rctl, rxcsum;
3183
3184         INIT_DEBUGOUT("em_initialize_receive_units: begin");
3185
3186         /*
3187          * Make sure receives are disabled while setting
3188          * up the descriptor ring
3189          */
3190         rctl = E1000_READ_REG(hw, E1000_RCTL);
3191         /* Do not disable if ever enabled on this hardware */
3192         if ((hw->mac.type != e1000_82574) && (hw->mac.type != e1000_82583))
3193                 E1000_WRITE_REG(hw, E1000_RCTL, rctl & ~E1000_RCTL_EN);
3194
3195         /* Setup the Receive Control Register */
3196         rctl &= ~(3 << E1000_RCTL_MO_SHIFT);
3197         rctl |= E1000_RCTL_EN | E1000_RCTL_BAM |
3198             E1000_RCTL_LBM_NO | E1000_RCTL_RDMTS_HALF |
3199             (hw->mac.mc_filter_type << E1000_RCTL_MO_SHIFT);
3200
3201         /* Do not store bad packets */
3202         rctl &= ~E1000_RCTL_SBP;
3203
3204         /* Enable Long Packet receive */
3205         if (if_getmtu(ifp) > ETHERMTU)
3206                 rctl |= E1000_RCTL_LPE;
3207         else
3208                 rctl &= ~E1000_RCTL_LPE;
3209
3210         /* Strip the CRC */
3211         if (!em_disable_crc_stripping)
3212                 rctl |= E1000_RCTL_SECRC;
3213
3214         if (hw->mac.type >= e1000_82540) {
3215                 E1000_WRITE_REG(hw, E1000_RADV,
3216                     sc->rx_abs_int_delay.value);
3217
3218                 /*
3219                  * Set the interrupt throttling rate. Value is calculated
3220                  * as DEFAULT_ITR = 1/(MAX_INTS_PER_SEC * 256ns)
3221                  */
3222                 E1000_WRITE_REG(hw, E1000_ITR, DEFAULT_ITR);
3223         }
3224         E1000_WRITE_REG(hw, E1000_RDTR, sc->rx_int_delay.value);
3225
3226         if (hw->mac.type >= em_mac_min) {
3227                 uint32_t rfctl;
3228                 /* Use extended rx descriptor formats */
3229                 rfctl = E1000_READ_REG(hw, E1000_RFCTL);
3230                 rfctl |= E1000_RFCTL_EXTEN;
3231
3232                 /*
3233                  * When using MSI-X interrupts we need to throttle
3234                  * using the EITR register (82574 only)
3235                  */
3236                 if (hw->mac.type == e1000_82574) {
3237                         for (int i = 0; i < 4; i++)
3238                                 E1000_WRITE_REG(hw, E1000_EITR_82574(i),
3239                                     DEFAULT_ITR);
3240                         /* Disable accelerated acknowledge */
3241                         rfctl |= E1000_RFCTL_ACK_DIS;
3242                 }
3243                 E1000_WRITE_REG(hw, E1000_RFCTL, rfctl);
3244         }
3245
3246         /* Set up L3 and L4 csum Rx descriptor offloads */
3247         rxcsum = E1000_READ_REG(hw, E1000_RXCSUM);
3248         if (if_getcapenable(ifp) & IFCAP_RXCSUM) {
3249                 rxcsum |= E1000_RXCSUM_TUOFL | E1000_RXCSUM_IPOFL;
3250                 if (hw->mac.type > e1000_82575)
3251                         rxcsum |= E1000_RXCSUM_CRCOFL;
3252                 else if (hw->mac.type < em_mac_min &&
3253                     if_getcapenable(ifp) & IFCAP_HWCSUM_IPV6)
3254                         rxcsum |= E1000_RXCSUM_IPV6OFL;
3255         } else {
3256                 rxcsum &= ~(E1000_RXCSUM_IPOFL | E1000_RXCSUM_TUOFL);
3257                 if (hw->mac.type > e1000_82575)
3258                         rxcsum &= ~E1000_RXCSUM_CRCOFL;
3259                 else if (hw->mac.type < em_mac_min)
3260                         rxcsum &= ~E1000_RXCSUM_IPV6OFL;
3261         }
3262
3263         if (sc->rx_num_queues > 1) {
3264                 /* RSS hash needed in the Rx descriptor */
3265                 rxcsum |= E1000_RXCSUM_PCSD;
3266
3267                 if (hw->mac.type >= igb_mac_min)
3268                         igb_initialize_rss_mapping(sc);
3269                 else
3270                         em_initialize_rss_mapping(sc);
3271         }
3272         E1000_WRITE_REG(hw, E1000_RXCSUM, rxcsum);
3273
3274         /*
3275          * XXX TEMPORARY WORKAROUND: on some systems with 82573
3276          * long latencies are observed, like Lenovo X60. This
3277          * change eliminates the problem, but since having positive
3278          * values in RDTR is a known source of problems on other
3279          * platforms another solution is being sought.
3280          */
3281         if (hw->mac.type == e1000_82573)
3282                 E1000_WRITE_REG(hw, E1000_RDTR, 0x20);
3283
3284         for (i = 0, que = sc->rx_queues; i < sc->rx_num_queues; i++, que++) {
3285                 struct rx_ring *rxr = &que->rxr;
3286                 /* Setup the Base and Length of the Rx Descriptor Ring */
3287                 u64 bus_addr = rxr->rx_paddr;
3288 #if 0
3289                 u32 rdt = sc->rx_num_queues -1;  /* default */
3290 #endif
3291
3292                 E1000_WRITE_REG(hw, E1000_RDLEN(i),
3293                     scctx->isc_nrxd[0] * sizeof(union e1000_rx_desc_extended));
3294                 E1000_WRITE_REG(hw, E1000_RDBAH(i), (u32)(bus_addr >> 32));
3295                 E1000_WRITE_REG(hw, E1000_RDBAL(i), (u32)bus_addr);
3296                 /* Setup the Head and Tail Descriptor Pointers */
3297                 E1000_WRITE_REG(hw, E1000_RDH(i), 0);
3298                 E1000_WRITE_REG(hw, E1000_RDT(i), 0);
3299         }
3300
3301         /*
3302          * Set PTHRESH for improved jumbo performance
3303          * According to 10.2.5.11 of Intel 82574 Datasheet,
3304          * RXDCTL(1) is written whenever RXDCTL(0) is written.
3305          * Only write to RXDCTL(1) if there is a need for different
3306          * settings.
3307          */
3308         if ((hw->mac.type == e1000_ich9lan || hw->mac.type == e1000_pch2lan ||
3309             hw->mac.type == e1000_ich10lan) && if_getmtu(ifp) > ETHERMTU) {
3310                 u32 rxdctl = E1000_READ_REG(hw, E1000_RXDCTL(0));
3311                 E1000_WRITE_REG(hw, E1000_RXDCTL(0), rxdctl | 3);
3312         } else if (hw->mac.type == e1000_82574) {
3313                 for (int i = 0; i < sc->rx_num_queues; i++) {
3314                         u32 rxdctl = E1000_READ_REG(hw, E1000_RXDCTL(i));
3315                         rxdctl |= 0x20; /* PTHRESH */
3316                         rxdctl |= 4 << 8; /* HTHRESH */
3317                         rxdctl |= 4 << 16;/* WTHRESH */
3318                         rxdctl |= 1 << 24; /* Switch to granularity */
3319                         E1000_WRITE_REG(hw, E1000_RXDCTL(i), rxdctl);
3320                 }
3321         } else if (hw->mac.type >= igb_mac_min) {
3322                 u32 psize, srrctl = 0;
3323
3324                 if (if_getmtu(ifp) > ETHERMTU) {
3325                         psize = scctx->isc_max_frame_size;
3326                         /* are we on a vlan? */
3327                         if (ifp->if_vlantrunk != NULL)
3328                                 psize += VLAN_TAG_SIZE;
3329
3330                         if (sc->vf_ifp)
3331                                 e1000_rlpml_set_vf(hw, psize);
3332                         else
3333                                 E1000_WRITE_REG(hw, E1000_RLPML, psize);
3334                 }
3335
3336                 /* Set maximum packet buffer len */
3337                 srrctl |= (sc->rx_mbuf_sz + BSIZEPKT_ROUNDUP) >>
3338                     E1000_SRRCTL_BSIZEPKT_SHIFT;
3339
3340                 /*
3341                  * If TX flow control is disabled and there's >1 queue defined,
3342                  * enable DROP.
3343                  *
3344                  * This drops frames rather than hanging the RX MAC for all queues.
3345                  */
3346                 if ((sc->rx_num_queues > 1) &&
3347                     (sc->fc == e1000_fc_none ||
3348                      sc->fc == e1000_fc_rx_pause)) {
3349                         srrctl |= E1000_SRRCTL_DROP_EN;
3350                 }
3351                         /* Setup the Base and Length of the Rx Descriptor Rings */
3352                 for (i = 0, que = sc->rx_queues; i < sc->rx_num_queues; i++, que++) {
3353                         struct rx_ring *rxr = &que->rxr;
3354                         u64 bus_addr = rxr->rx_paddr;
3355                         u32 rxdctl;
3356
3357 #ifdef notyet
3358                         /* Configure for header split? -- ignore for now */
3359                         rxr->hdr_split = igb_header_split;
3360 #else
3361                         srrctl |= E1000_SRRCTL_DESCTYPE_ADV_ONEBUF;
3362 #endif
3363
3364                         E1000_WRITE_REG(hw, E1000_RDLEN(i),
3365                                         scctx->isc_nrxd[0] * sizeof(struct e1000_rx_desc));
3366                         E1000_WRITE_REG(hw, E1000_RDBAH(i),
3367                                         (uint32_t)(bus_addr >> 32));
3368                         E1000_WRITE_REG(hw, E1000_RDBAL(i),
3369                                         (uint32_t)bus_addr);
3370                         E1000_WRITE_REG(hw, E1000_SRRCTL(i), srrctl);
3371                         /* Enable this Queue */
3372                         rxdctl = E1000_READ_REG(hw, E1000_RXDCTL(i));
3373                         rxdctl |= E1000_RXDCTL_QUEUE_ENABLE;
3374                         rxdctl &= 0xFFF00000;
3375                         rxdctl |= IGB_RX_PTHRESH;
3376                         rxdctl |= IGB_RX_HTHRESH << 8;
3377                         rxdctl |= IGB_RX_WTHRESH << 16;
3378                         E1000_WRITE_REG(hw, E1000_RXDCTL(i), rxdctl);
3379                 }               
3380         } else if (hw->mac.type >= e1000_pch2lan) {
3381                 if (if_getmtu(ifp) > ETHERMTU)
3382                         e1000_lv_jumbo_workaround_ich8lan(hw, true);
3383                 else
3384                         e1000_lv_jumbo_workaround_ich8lan(hw, false);
3385         }
3386
3387         /* Make sure VLAN Filters are off */
3388         rctl &= ~E1000_RCTL_VFE;
3389
3390         /* Set up packet buffer size, overridden by per queue srrctl on igb */
3391         if (hw->mac.type < igb_mac_min) {
3392                 if (sc->rx_mbuf_sz > 2048 && sc->rx_mbuf_sz <= 4096)
3393                         rctl |= E1000_RCTL_SZ_4096 | E1000_RCTL_BSEX;
3394                 else if (sc->rx_mbuf_sz > 4096 && sc->rx_mbuf_sz <= 8192)
3395                         rctl |= E1000_RCTL_SZ_8192 | E1000_RCTL_BSEX;
3396                 else if (sc->rx_mbuf_sz > 8192)
3397                         rctl |= E1000_RCTL_SZ_16384 | E1000_RCTL_BSEX;
3398                 else {
3399                         rctl |= E1000_RCTL_SZ_2048;
3400                         rctl &= ~E1000_RCTL_BSEX;
3401                 }
3402         } else
3403                 rctl |= E1000_RCTL_SZ_2048;
3404
3405         /*
3406          * rctl bits 11:10 are as follows
3407          * lem: reserved
3408          * em: DTYPE
3409          * igb: reserved
3410          * and should be 00 on all of the above
3411          */
3412         rctl &= ~0x00000C00;
3413
3414         /* Write out the settings */
3415         E1000_WRITE_REG(hw, E1000_RCTL, rctl);
3416
3417         return;
3418 }
3419
3420 static void
3421 em_if_vlan_register(if_ctx_t ctx, u16 vtag)
3422 {
3423         struct e1000_softc *sc = iflib_get_softc(ctx);
3424         u32 index, bit;
3425
3426         index = (vtag >> 5) & 0x7F;
3427         bit = vtag & 0x1F;
3428         sc->shadow_vfta[index] |= (1 << bit);
3429         ++sc->num_vlans;
3430         em_if_vlan_filter_write(sc);
3431 }
3432
3433 static void
3434 em_if_vlan_unregister(if_ctx_t ctx, u16 vtag)
3435 {
3436         struct e1000_softc *sc = iflib_get_softc(ctx);
3437         u32 index, bit;
3438
3439         index = (vtag >> 5) & 0x7F;
3440         bit = vtag & 0x1F;
3441         sc->shadow_vfta[index] &= ~(1 << bit);
3442         --sc->num_vlans;
3443         em_if_vlan_filter_write(sc);
3444 }
3445
3446 static bool
3447 em_if_vlan_filter_capable(if_ctx_t ctx)
3448 {
3449         if_t ifp = iflib_get_ifp(ctx);
3450
3451         if ((if_getcapenable(ifp) & IFCAP_VLAN_HWFILTER) &&
3452             !em_disable_crc_stripping)
3453                 return (true);
3454
3455         return (false);
3456 }
3457
3458 static bool
3459 em_if_vlan_filter_used(if_ctx_t ctx)
3460 {
3461         struct e1000_softc *sc = iflib_get_softc(ctx);
3462
3463         if (!em_if_vlan_filter_capable(ctx))
3464                 return (false);
3465
3466         for (int i = 0; i < EM_VFTA_SIZE; i++)
3467                 if (sc->shadow_vfta[i] != 0)
3468                         return (true);
3469
3470         return (false);
3471 }
3472
3473 static void
3474 em_if_vlan_filter_enable(struct e1000_softc *sc)
3475 {
3476         struct e1000_hw *hw = &sc->hw;
3477         u32 reg;
3478
3479         reg = E1000_READ_REG(hw, E1000_RCTL);
3480         reg &= ~E1000_RCTL_CFIEN;
3481         reg |= E1000_RCTL_VFE;
3482         E1000_WRITE_REG(hw, E1000_RCTL, reg);
3483 }
3484
3485 static void
3486 em_if_vlan_filter_disable(struct e1000_softc *sc)
3487 {
3488         struct e1000_hw *hw = &sc->hw;
3489         u32 reg;
3490
3491         reg = E1000_READ_REG(hw, E1000_RCTL);
3492         reg &= ~(E1000_RCTL_VFE | E1000_RCTL_CFIEN);
3493         E1000_WRITE_REG(hw, E1000_RCTL, reg);
3494 }
3495
3496 static void
3497 em_if_vlan_filter_write(struct e1000_softc *sc)
3498 {
3499         struct e1000_hw *hw = &sc->hw;
3500
3501         if (sc->vf_ifp)
3502                 return;
3503
3504         /* Disable interrupts for lem-class devices during the filter change */
3505         if (hw->mac.type < em_mac_min)
3506                 em_if_intr_disable(sc->ctx);
3507
3508         for (int i = 0; i < EM_VFTA_SIZE; i++)
3509                 if (sc->shadow_vfta[i] != 0) {
3510                         /* XXXKB: incomplete VF support, we return early above */
3511                         if (sc->vf_ifp)
3512                                 e1000_vfta_set_vf(hw, sc->shadow_vfta[i], true);
3513                         else
3514                                 e1000_write_vfta(hw, i, sc->shadow_vfta[i]);
3515                 }
3516
3517         /* Re-enable interrupts for lem-class devices */
3518         if (hw->mac.type < em_mac_min)
3519                 em_if_intr_enable(sc->ctx);
3520 }
3521
3522 static void
3523 em_setup_vlan_hw_support(if_ctx_t ctx)
3524 {
3525         struct e1000_softc *sc = iflib_get_softc(ctx);
3526         struct e1000_hw *hw = &sc->hw;
3527         struct ifnet *ifp = iflib_get_ifp(ctx);
3528         u32 reg;
3529
3530         /* XXXKB: Return early if we are a VF until VF decap and filter management
3531          * is ready and tested.
3532          */
3533         if (sc->vf_ifp)
3534                 return;
3535
3536         if (if_getcapenable(ifp) & IFCAP_VLAN_HWTAGGING &&
3537             !em_disable_crc_stripping) {
3538                 reg = E1000_READ_REG(hw, E1000_CTRL);
3539                 reg |= E1000_CTRL_VME;
3540                 E1000_WRITE_REG(hw, E1000_CTRL, reg);
3541         } else {
3542                 reg = E1000_READ_REG(hw, E1000_CTRL);
3543                 reg &= ~E1000_CTRL_VME;
3544                 E1000_WRITE_REG(hw, E1000_CTRL, reg);
3545         }
3546
3547         /* If we aren't doing HW filtering, we're done */
3548         if (!em_if_vlan_filter_capable(ctx))  {
3549                 em_if_vlan_filter_disable(sc);
3550                 return;
3551         }
3552
3553         /*
3554          * A soft reset zero's out the VFTA, so
3555          * we need to repopulate it now.
3556          */
3557         em_if_vlan_filter_write(sc);
3558
3559         /* Enable the Filter Table */
3560         em_if_vlan_filter_enable(sc);
3561 }
3562
3563 static void
3564 em_if_intr_enable(if_ctx_t ctx)
3565 {
3566         struct e1000_softc *sc = iflib_get_softc(ctx);
3567         struct e1000_hw *hw = &sc->hw;
3568         u32 ims_mask = IMS_ENABLE_MASK;
3569
3570         if (sc->intr_type == IFLIB_INTR_MSIX) {
3571                 E1000_WRITE_REG(hw, EM_EIAC, sc->ims);
3572                 ims_mask |= sc->ims;
3573         }
3574         E1000_WRITE_REG(hw, E1000_IMS, ims_mask);
3575         E1000_WRITE_FLUSH(hw);
3576 }
3577
3578 static void
3579 em_if_intr_disable(if_ctx_t ctx)
3580 {
3581         struct e1000_softc *sc = iflib_get_softc(ctx);
3582         struct e1000_hw *hw = &sc->hw;
3583
3584         if (sc->intr_type == IFLIB_INTR_MSIX)
3585                 E1000_WRITE_REG(hw, EM_EIAC, 0);
3586         E1000_WRITE_REG(hw, E1000_IMC, 0xffffffff);
3587         E1000_WRITE_FLUSH(hw);
3588 }
3589
3590 static void
3591 igb_if_intr_enable(if_ctx_t ctx)
3592 {
3593         struct e1000_softc *sc = iflib_get_softc(ctx);
3594         struct e1000_hw *hw = &sc->hw;
3595         u32 mask;
3596
3597         if (__predict_true(sc->intr_type == IFLIB_INTR_MSIX)) {
3598                 mask = (sc->que_mask | sc->link_mask);
3599                 E1000_WRITE_REG(hw, E1000_EIAC, mask);
3600                 E1000_WRITE_REG(hw, E1000_EIAM, mask);
3601                 E1000_WRITE_REG(hw, E1000_EIMS, mask);
3602                 E1000_WRITE_REG(hw, E1000_IMS, E1000_IMS_LSC);
3603         } else
3604                 E1000_WRITE_REG(hw, E1000_IMS, IMS_ENABLE_MASK);
3605         E1000_WRITE_FLUSH(hw);
3606 }
3607
3608 static void
3609 igb_if_intr_disable(if_ctx_t ctx)
3610 {
3611         struct e1000_softc *sc = iflib_get_softc(ctx);
3612         struct e1000_hw *hw = &sc->hw;
3613
3614         if (__predict_true(sc->intr_type == IFLIB_INTR_MSIX)) {
3615                 E1000_WRITE_REG(hw, E1000_EIMC, 0xffffffff);
3616                 E1000_WRITE_REG(hw, E1000_EIAC, 0);
3617         }
3618         E1000_WRITE_REG(hw, E1000_IMC, 0xffffffff);
3619         E1000_WRITE_FLUSH(hw);
3620 }
3621
3622 /*
3623  * Bit of a misnomer, what this really means is
3624  * to enable OS management of the system... aka
3625  * to disable special hardware management features
3626  */
3627 static void
3628 em_init_manageability(struct e1000_softc *sc)
3629 {
3630         /* A shared code workaround */
3631 #define E1000_82542_MANC2H E1000_MANC2H
3632         if (sc->has_manage) {
3633                 int manc2h = E1000_READ_REG(&sc->hw, E1000_MANC2H);
3634                 int manc = E1000_READ_REG(&sc->hw, E1000_MANC);
3635
3636                 /* disable hardware interception of ARP */
3637                 manc &= ~(E1000_MANC_ARP_EN);
3638
3639                 /* enable receiving management packets to the host */
3640                 manc |= E1000_MANC_EN_MNG2HOST;
3641 #define E1000_MNG2HOST_PORT_623 (1 << 5)
3642 #define E1000_MNG2HOST_PORT_664 (1 << 6)
3643                 manc2h |= E1000_MNG2HOST_PORT_623;
3644                 manc2h |= E1000_MNG2HOST_PORT_664;
3645                 E1000_WRITE_REG(&sc->hw, E1000_MANC2H, manc2h);
3646                 E1000_WRITE_REG(&sc->hw, E1000_MANC, manc);
3647         }
3648 }
3649
3650 /*
3651  * Give control back to hardware management
3652  * controller if there is one.
3653  */
3654 static void
3655 em_release_manageability(struct e1000_softc *sc)
3656 {
3657         if (sc->has_manage) {
3658                 int manc = E1000_READ_REG(&sc->hw, E1000_MANC);
3659
3660                 /* re-enable hardware interception of ARP */
3661                 manc |= E1000_MANC_ARP_EN;
3662                 manc &= ~E1000_MANC_EN_MNG2HOST;
3663
3664                 E1000_WRITE_REG(&sc->hw, E1000_MANC, manc);
3665         }
3666 }
3667
3668 /*
3669  * em_get_hw_control sets the {CTRL_EXT|FWSM}:DRV_LOAD bit.
3670  * For ASF and Pass Through versions of f/w this means
3671  * that the driver is loaded. For AMT version type f/w
3672  * this means that the network i/f is open.
3673  */
3674 static void
3675 em_get_hw_control(struct e1000_softc *sc)
3676 {
3677         u32 ctrl_ext, swsm;
3678
3679         if (sc->vf_ifp)
3680                 return;
3681
3682         if (sc->hw.mac.type == e1000_82573) {
3683                 swsm = E1000_READ_REG(&sc->hw, E1000_SWSM);
3684                 E1000_WRITE_REG(&sc->hw, E1000_SWSM,
3685                     swsm | E1000_SWSM_DRV_LOAD);
3686                 return;
3687         }
3688         /* else */
3689         ctrl_ext = E1000_READ_REG(&sc->hw, E1000_CTRL_EXT);
3690         E1000_WRITE_REG(&sc->hw, E1000_CTRL_EXT,
3691             ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
3692 }
3693
3694 /*
3695  * em_release_hw_control resets {CTRL_EXT|FWSM}:DRV_LOAD bit.
3696  * For ASF and Pass Through versions of f/w this means that
3697  * the driver is no longer loaded. For AMT versions of the
3698  * f/w this means that the network i/f is closed.
3699  */
3700 static void
3701 em_release_hw_control(struct e1000_softc *sc)
3702 {
3703         u32 ctrl_ext, swsm;
3704
3705         if (!sc->has_manage)
3706                 return;
3707
3708         if (sc->hw.mac.type == e1000_82573) {
3709                 swsm = E1000_READ_REG(&sc->hw, E1000_SWSM);
3710                 E1000_WRITE_REG(&sc->hw, E1000_SWSM,
3711                     swsm & ~E1000_SWSM_DRV_LOAD);
3712                 return;
3713         }
3714         /* else */
3715         ctrl_ext = E1000_READ_REG(&sc->hw, E1000_CTRL_EXT);
3716         E1000_WRITE_REG(&sc->hw, E1000_CTRL_EXT,
3717             ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
3718         return;
3719 }
3720
3721 static int
3722 em_is_valid_ether_addr(u8 *addr)
3723 {
3724         char zero_addr[6] = { 0, 0, 0, 0, 0, 0 };
3725
3726         if ((addr[0] & 1) || (!bcmp(addr, zero_addr, ETHER_ADDR_LEN))) {
3727                 return (false);
3728         }
3729
3730         return (true);
3731 }
3732
3733 /*
3734 ** Parse the interface capabilities with regard
3735 ** to both system management and wake-on-lan for
3736 ** later use.
3737 */
3738 static void
3739 em_get_wakeup(if_ctx_t ctx)
3740 {
3741         struct e1000_softc *sc = iflib_get_softc(ctx);
3742         device_t dev = iflib_get_dev(ctx);
3743         u16 eeprom_data = 0, device_id, apme_mask;
3744
3745         sc->has_manage = e1000_enable_mng_pass_thru(&sc->hw);
3746         apme_mask = EM_EEPROM_APME;
3747
3748         switch (sc->hw.mac.type) {
3749         case e1000_82542:
3750         case e1000_82543:
3751                 break;
3752         case e1000_82544:
3753                 e1000_read_nvm(&sc->hw,
3754                     NVM_INIT_CONTROL2_REG, 1, &eeprom_data);
3755                 apme_mask = EM_82544_APME;
3756                 break;
3757         case e1000_82546:
3758         case e1000_82546_rev_3:
3759                 if (sc->hw.bus.func == 1) {
3760                         e1000_read_nvm(&sc->hw,
3761                             NVM_INIT_CONTROL3_PORT_B, 1, &eeprom_data);
3762                         break;
3763                 } else
3764                         e1000_read_nvm(&sc->hw,
3765                             NVM_INIT_CONTROL3_PORT_A, 1, &eeprom_data);
3766                 break;
3767         case e1000_82573:
3768         case e1000_82583:
3769                 sc->has_amt = true;
3770                 /* FALLTHROUGH */
3771         case e1000_82571:
3772         case e1000_82572:
3773         case e1000_80003es2lan:
3774                 if (sc->hw.bus.func == 1) {
3775                         e1000_read_nvm(&sc->hw,
3776                             NVM_INIT_CONTROL3_PORT_B, 1, &eeprom_data);
3777                         break;
3778                 } else
3779                         e1000_read_nvm(&sc->hw,
3780                             NVM_INIT_CONTROL3_PORT_A, 1, &eeprom_data);
3781                 break;
3782         case e1000_ich8lan:
3783         case e1000_ich9lan:
3784         case e1000_ich10lan:
3785         case e1000_pchlan:
3786         case e1000_pch2lan:
3787         case e1000_pch_lpt:
3788         case e1000_pch_spt:
3789         case e1000_82575:       /* listing all igb devices */
3790         case e1000_82576:
3791         case e1000_82580:
3792         case e1000_i350:
3793         case e1000_i354:
3794         case e1000_i210:
3795         case e1000_i211:
3796         case e1000_vfadapt:
3797         case e1000_vfadapt_i350:
3798                 apme_mask = E1000_WUC_APME;
3799                 sc->has_amt = true;
3800                 eeprom_data = E1000_READ_REG(&sc->hw, E1000_WUC);
3801                 break;
3802         default:
3803                 e1000_read_nvm(&sc->hw,
3804                     NVM_INIT_CONTROL3_PORT_A, 1, &eeprom_data);
3805                 break;
3806         }
3807         if (eeprom_data & apme_mask)
3808                 sc->wol = (E1000_WUFC_MAG | E1000_WUFC_MC);
3809         /*
3810          * We have the eeprom settings, now apply the special cases
3811          * where the eeprom may be wrong or the board won't support
3812          * wake on lan on a particular port
3813          */
3814         device_id = pci_get_device(dev);
3815         switch (device_id) {
3816         case E1000_DEV_ID_82546GB_PCIE:
3817                 sc->wol = 0;
3818                 break;
3819         case E1000_DEV_ID_82546EB_FIBER:
3820         case E1000_DEV_ID_82546GB_FIBER:
3821                 /* Wake events only supported on port A for dual fiber
3822                  * regardless of eeprom setting */
3823                 if (E1000_READ_REG(&sc->hw, E1000_STATUS) &
3824                     E1000_STATUS_FUNC_1)
3825                         sc->wol = 0;
3826                 break;
3827         case E1000_DEV_ID_82546GB_QUAD_COPPER_KSP3:
3828                 /* if quad port adapter, disable WoL on all but port A */
3829                 if (global_quad_port_a != 0)
3830                         sc->wol = 0;
3831                 /* Reset for multiple quad port adapters */
3832                 if (++global_quad_port_a == 4)
3833                         global_quad_port_a = 0;
3834                 break;
3835         case E1000_DEV_ID_82571EB_FIBER:
3836                 /* Wake events only supported on port A for dual fiber
3837                  * regardless of eeprom setting */
3838                 if (E1000_READ_REG(&sc->hw, E1000_STATUS) &
3839                     E1000_STATUS_FUNC_1)
3840                         sc->wol = 0;
3841                 break;
3842         case E1000_DEV_ID_82571EB_QUAD_COPPER:
3843         case E1000_DEV_ID_82571EB_QUAD_FIBER:
3844         case E1000_DEV_ID_82571EB_QUAD_COPPER_LP:
3845                 /* if quad port adapter, disable WoL on all but port A */
3846                 if (global_quad_port_a != 0)
3847                         sc->wol = 0;
3848                 /* Reset for multiple quad port adapters */
3849                 if (++global_quad_port_a == 4)
3850                         global_quad_port_a = 0;
3851                 break;
3852         }
3853         return;
3854 }
3855
3856
3857 /*
3858  * Enable PCI Wake On Lan capability
3859  */
3860 static void
3861 em_enable_wakeup(if_ctx_t ctx)
3862 {
3863         struct e1000_softc *sc = iflib_get_softc(ctx);
3864         device_t dev = iflib_get_dev(ctx);
3865         if_t ifp = iflib_get_ifp(ctx);
3866         int error = 0;
3867         u32 pmc, ctrl, ctrl_ext, rctl;
3868         u16 status;
3869
3870         if (pci_find_cap(dev, PCIY_PMG, &pmc) != 0)
3871                 return;
3872
3873         /*
3874          * Determine type of Wakeup: note that wol
3875          * is set with all bits on by default.
3876          */
3877         if ((if_getcapenable(ifp) & IFCAP_WOL_MAGIC) == 0)
3878                 sc->wol &= ~E1000_WUFC_MAG;
3879
3880         if ((if_getcapenable(ifp) & IFCAP_WOL_UCAST) == 0)
3881                 sc->wol &= ~E1000_WUFC_EX;
3882
3883         if ((if_getcapenable(ifp) & IFCAP_WOL_MCAST) == 0)
3884                 sc->wol &= ~E1000_WUFC_MC;
3885         else {
3886                 rctl = E1000_READ_REG(&sc->hw, E1000_RCTL);
3887                 rctl |= E1000_RCTL_MPE;
3888                 E1000_WRITE_REG(&sc->hw, E1000_RCTL, rctl);
3889         }
3890
3891         if (!(sc->wol & (E1000_WUFC_EX | E1000_WUFC_MAG | E1000_WUFC_MC)))
3892                 goto pme;
3893
3894         /* Advertise the wakeup capability */
3895         ctrl = E1000_READ_REG(&sc->hw, E1000_CTRL);
3896         ctrl |= (E1000_CTRL_SWDPIN2 | E1000_CTRL_SWDPIN3);
3897         E1000_WRITE_REG(&sc->hw, E1000_CTRL, ctrl);
3898
3899         /* Keep the laser running on Fiber adapters */
3900         if (sc->hw.phy.media_type == e1000_media_type_fiber ||
3901             sc->hw.phy.media_type == e1000_media_type_internal_serdes) {
3902                 ctrl_ext = E1000_READ_REG(&sc->hw, E1000_CTRL_EXT);
3903                 ctrl_ext |= E1000_CTRL_EXT_SDP3_DATA;
3904                 E1000_WRITE_REG(&sc->hw, E1000_CTRL_EXT, ctrl_ext);
3905         }
3906
3907         if ((sc->hw.mac.type == e1000_ich8lan) ||
3908             (sc->hw.mac.type == e1000_pchlan) ||
3909             (sc->hw.mac.type == e1000_ich9lan) ||
3910             (sc->hw.mac.type == e1000_ich10lan))
3911                 e1000_suspend_workarounds_ich8lan(&sc->hw);
3912
3913         if ( sc->hw.mac.type >= e1000_pchlan) {
3914                 error = em_enable_phy_wakeup(sc);
3915                 if (error)
3916                         goto pme;
3917         } else {
3918                 /* Enable wakeup by the MAC */
3919                 E1000_WRITE_REG(&sc->hw, E1000_WUC, E1000_WUC_PME_EN);
3920                 E1000_WRITE_REG(&sc->hw, E1000_WUFC, sc->wol);
3921         }
3922
3923         if (sc->hw.phy.type == e1000_phy_igp_3)
3924                 e1000_igp3_phy_powerdown_workaround_ich8lan(&sc->hw);
3925
3926 pme:
3927         status = pci_read_config(dev, pmc + PCIR_POWER_STATUS, 2);
3928         status &= ~(PCIM_PSTAT_PME | PCIM_PSTAT_PMEENABLE);
3929         if (!error && (if_getcapenable(ifp) & IFCAP_WOL))
3930                 status |= PCIM_PSTAT_PME | PCIM_PSTAT_PMEENABLE;
3931         pci_write_config(dev, pmc + PCIR_POWER_STATUS, status, 2);
3932
3933         return;
3934 }
3935
3936 /*
3937  * WOL in the newer chipset interfaces (pchlan)
3938  * require thing to be copied into the phy
3939  */
3940 static int
3941 em_enable_phy_wakeup(struct e1000_softc *sc)
3942 {
3943         struct e1000_hw *hw = &sc->hw;
3944         u32 mreg, ret = 0;
3945         u16 preg;
3946
3947         /* copy MAC RARs to PHY RARs */
3948         e1000_copy_rx_addrs_to_phy_ich8lan(hw);
3949
3950         /* copy MAC MTA to PHY MTA */
3951         for (int i = 0; i < hw->mac.mta_reg_count; i++) {
3952                 mreg = E1000_READ_REG_ARRAY(hw, E1000_MTA, i);
3953                 e1000_write_phy_reg(hw, BM_MTA(i), (u16)(mreg & 0xFFFF));
3954                 e1000_write_phy_reg(hw, BM_MTA(i) + 1,
3955                     (u16)((mreg >> 16) & 0xFFFF));
3956         }
3957
3958         /* configure PHY Rx Control register */
3959         e1000_read_phy_reg(hw, BM_RCTL, &preg);
3960         mreg = E1000_READ_REG(hw, E1000_RCTL);
3961         if (mreg & E1000_RCTL_UPE)
3962                 preg |= BM_RCTL_UPE;
3963         if (mreg & E1000_RCTL_MPE)
3964                 preg |= BM_RCTL_MPE;
3965         preg &= ~(BM_RCTL_MO_MASK);
3966         if (mreg & E1000_RCTL_MO_3)
3967                 preg |= (((mreg & E1000_RCTL_MO_3) >> E1000_RCTL_MO_SHIFT)
3968                                 << BM_RCTL_MO_SHIFT);
3969         if (mreg & E1000_RCTL_BAM)
3970                 preg |= BM_RCTL_BAM;
3971         if (mreg & E1000_RCTL_PMCF)
3972                 preg |= BM_RCTL_PMCF;
3973         mreg = E1000_READ_REG(hw, E1000_CTRL);
3974         if (mreg & E1000_CTRL_RFCE)
3975                 preg |= BM_RCTL_RFCE;
3976         e1000_write_phy_reg(hw, BM_RCTL, preg);
3977
3978         /* enable PHY wakeup in MAC register */
3979         E1000_WRITE_REG(hw, E1000_WUC,
3980             E1000_WUC_PHY_WAKE | E1000_WUC_PME_EN | E1000_WUC_APME);
3981         E1000_WRITE_REG(hw, E1000_WUFC, sc->wol);
3982
3983         /* configure and enable PHY wakeup in PHY registers */
3984         e1000_write_phy_reg(hw, BM_WUFC, sc->wol);
3985         e1000_write_phy_reg(hw, BM_WUC, E1000_WUC_PME_EN);
3986
3987         /* activate PHY wakeup */
3988         ret = hw->phy.ops.acquire(hw);
3989         if (ret) {
3990                 printf("Could not acquire PHY\n");
3991                 return ret;
3992         }
3993         e1000_write_phy_reg_mdic(hw, IGP01E1000_PHY_PAGE_SELECT,
3994                                  (BM_WUC_ENABLE_PAGE << IGP_PAGE_SHIFT));
3995         ret = e1000_read_phy_reg_mdic(hw, BM_WUC_ENABLE_REG, &preg);
3996         if (ret) {
3997                 printf("Could not read PHY page 769\n");
3998                 goto out;
3999         }
4000         preg |= BM_WUC_ENABLE_BIT | BM_WUC_HOST_WU_BIT;
4001         ret = e1000_write_phy_reg_mdic(hw, BM_WUC_ENABLE_REG, preg);
4002         if (ret)
4003                 printf("Could not set PHY Host Wakeup bit\n");
4004 out:
4005         hw->phy.ops.release(hw);
4006
4007         return ret;
4008 }
4009
4010 static void
4011 em_if_led_func(if_ctx_t ctx, int onoff)
4012 {
4013         struct e1000_softc *sc = iflib_get_softc(ctx);
4014
4015         if (onoff) {
4016                 e1000_setup_led(&sc->hw);
4017                 e1000_led_on(&sc->hw);
4018         } else {
4019                 e1000_led_off(&sc->hw);
4020                 e1000_cleanup_led(&sc->hw);
4021         }
4022 }
4023
4024 /*
4025  * Disable the L0S and L1 LINK states
4026  */
4027 static void
4028 em_disable_aspm(struct e1000_softc *sc)
4029 {
4030         int base, reg;
4031         u16 link_cap,link_ctrl;
4032         device_t dev = sc->dev;
4033
4034         switch (sc->hw.mac.type) {
4035         case e1000_82573:
4036         case e1000_82574:
4037         case e1000_82583:
4038                 break;
4039         default:
4040                 return;
4041         }
4042         if (pci_find_cap(dev, PCIY_EXPRESS, &base) != 0)
4043                 return;
4044         reg = base + PCIER_LINK_CAP;
4045         link_cap = pci_read_config(dev, reg, 2);
4046         if ((link_cap & PCIEM_LINK_CAP_ASPM) == 0)
4047                 return;
4048         reg = base + PCIER_LINK_CTL;
4049         link_ctrl = pci_read_config(dev, reg, 2);
4050         link_ctrl &= ~PCIEM_LINK_CTL_ASPMC;
4051         pci_write_config(dev, reg, link_ctrl, 2);
4052         return;
4053 }
4054
4055 /**********************************************************************
4056  *
4057  *  Update the board statistics counters.
4058  *
4059  **********************************************************************/
4060 static void
4061 em_update_stats_counters(struct e1000_softc *sc)
4062 {
4063         u64 prev_xoffrxc = sc->stats.xoffrxc;
4064
4065         if(sc->hw.phy.media_type == e1000_media_type_copper ||
4066            (E1000_READ_REG(&sc->hw, E1000_STATUS) & E1000_STATUS_LU)) {
4067                 sc->stats.symerrs += E1000_READ_REG(&sc->hw, E1000_SYMERRS);
4068                 sc->stats.sec += E1000_READ_REG(&sc->hw, E1000_SEC);
4069         }
4070         sc->stats.crcerrs += E1000_READ_REG(&sc->hw, E1000_CRCERRS);
4071         sc->stats.mpc += E1000_READ_REG(&sc->hw, E1000_MPC);
4072         sc->stats.scc += E1000_READ_REG(&sc->hw, E1000_SCC);
4073         sc->stats.ecol += E1000_READ_REG(&sc->hw, E1000_ECOL);
4074
4075         sc->stats.mcc += E1000_READ_REG(&sc->hw, E1000_MCC);
4076         sc->stats.latecol += E1000_READ_REG(&sc->hw, E1000_LATECOL);
4077         sc->stats.colc += E1000_READ_REG(&sc->hw, E1000_COLC);
4078         sc->stats.dc += E1000_READ_REG(&sc->hw, E1000_DC);
4079         sc->stats.rlec += E1000_READ_REG(&sc->hw, E1000_RLEC);
4080         sc->stats.xonrxc += E1000_READ_REG(&sc->hw, E1000_XONRXC);
4081         sc->stats.xontxc += E1000_READ_REG(&sc->hw, E1000_XONTXC);
4082         sc->stats.xoffrxc += E1000_READ_REG(&sc->hw, E1000_XOFFRXC);
4083         /*
4084          ** For watchdog management we need to know if we have been
4085          ** paused during the last interval, so capture that here.
4086         */
4087         if (sc->stats.xoffrxc != prev_xoffrxc)
4088                 sc->shared->isc_pause_frames = 1;
4089         sc->stats.xofftxc += E1000_READ_REG(&sc->hw, E1000_XOFFTXC);
4090         sc->stats.fcruc += E1000_READ_REG(&sc->hw, E1000_FCRUC);
4091         sc->stats.prc64 += E1000_READ_REG(&sc->hw, E1000_PRC64);
4092         sc->stats.prc127 += E1000_READ_REG(&sc->hw, E1000_PRC127);
4093         sc->stats.prc255 += E1000_READ_REG(&sc->hw, E1000_PRC255);
4094         sc->stats.prc511 += E1000_READ_REG(&sc->hw, E1000_PRC511);
4095         sc->stats.prc1023 += E1000_READ_REG(&sc->hw, E1000_PRC1023);
4096         sc->stats.prc1522 += E1000_READ_REG(&sc->hw, E1000_PRC1522);
4097         sc->stats.gprc += E1000_READ_REG(&sc->hw, E1000_GPRC);
4098         sc->stats.bprc += E1000_READ_REG(&sc->hw, E1000_BPRC);
4099         sc->stats.mprc += E1000_READ_REG(&sc->hw, E1000_MPRC);
4100         sc->stats.gptc += E1000_READ_REG(&sc->hw, E1000_GPTC);
4101
4102         /* For the 64-bit byte counters the low dword must be read first. */
4103         /* Both registers clear on the read of the high dword */
4104
4105         sc->stats.gorc += E1000_READ_REG(&sc->hw, E1000_GORCL) +
4106             ((u64)E1000_READ_REG(&sc->hw, E1000_GORCH) << 32);
4107         sc->stats.gotc += E1000_READ_REG(&sc->hw, E1000_GOTCL) +
4108             ((u64)E1000_READ_REG(&sc->hw, E1000_GOTCH) << 32);
4109
4110         sc->stats.rnbc += E1000_READ_REG(&sc->hw, E1000_RNBC);
4111         sc->stats.ruc += E1000_READ_REG(&sc->hw, E1000_RUC);
4112         sc->stats.rfc += E1000_READ_REG(&sc->hw, E1000_RFC);
4113         sc->stats.roc += E1000_READ_REG(&sc->hw, E1000_ROC);
4114         sc->stats.rjc += E1000_READ_REG(&sc->hw, E1000_RJC);
4115
4116         sc->stats.tor += E1000_READ_REG(&sc->hw, E1000_TORH);
4117         sc->stats.tot += E1000_READ_REG(&sc->hw, E1000_TOTH);
4118
4119         sc->stats.tpr += E1000_READ_REG(&sc->hw, E1000_TPR);
4120         sc->stats.tpt += E1000_READ_REG(&sc->hw, E1000_TPT);
4121         sc->stats.ptc64 += E1000_READ_REG(&sc->hw, E1000_PTC64);
4122         sc->stats.ptc127 += E1000_READ_REG(&sc->hw, E1000_PTC127);
4123         sc->stats.ptc255 += E1000_READ_REG(&sc->hw, E1000_PTC255);
4124         sc->stats.ptc511 += E1000_READ_REG(&sc->hw, E1000_PTC511);
4125         sc->stats.ptc1023 += E1000_READ_REG(&sc->hw, E1000_PTC1023);
4126         sc->stats.ptc1522 += E1000_READ_REG(&sc->hw, E1000_PTC1522);
4127         sc->stats.mptc += E1000_READ_REG(&sc->hw, E1000_MPTC);
4128         sc->stats.bptc += E1000_READ_REG(&sc->hw, E1000_BPTC);
4129
4130         /* Interrupt Counts */
4131
4132         sc->stats.iac += E1000_READ_REG(&sc->hw, E1000_IAC);
4133         sc->stats.icrxptc += E1000_READ_REG(&sc->hw, E1000_ICRXPTC);
4134         sc->stats.icrxatc += E1000_READ_REG(&sc->hw, E1000_ICRXATC);
4135         sc->stats.ictxptc += E1000_READ_REG(&sc->hw, E1000_ICTXPTC);
4136         sc->stats.ictxatc += E1000_READ_REG(&sc->hw, E1000_ICTXATC);
4137         sc->stats.ictxqec += E1000_READ_REG(&sc->hw, E1000_ICTXQEC);
4138         sc->stats.ictxqmtc += E1000_READ_REG(&sc->hw, E1000_ICTXQMTC);
4139         sc->stats.icrxdmtc += E1000_READ_REG(&sc->hw, E1000_ICRXDMTC);
4140         sc->stats.icrxoc += E1000_READ_REG(&sc->hw, E1000_ICRXOC);
4141
4142         if (sc->hw.mac.type >= e1000_82543) {
4143                 sc->stats.algnerrc +=
4144                 E1000_READ_REG(&sc->hw, E1000_ALGNERRC);
4145                 sc->stats.rxerrc +=
4146                 E1000_READ_REG(&sc->hw, E1000_RXERRC);
4147                 sc->stats.tncrs +=
4148                 E1000_READ_REG(&sc->hw, E1000_TNCRS);
4149                 sc->stats.cexterr +=
4150                 E1000_READ_REG(&sc->hw, E1000_CEXTERR);
4151                 sc->stats.tsctc +=
4152                 E1000_READ_REG(&sc->hw, E1000_TSCTC);
4153                 sc->stats.tsctfc +=
4154                 E1000_READ_REG(&sc->hw, E1000_TSCTFC);
4155         }
4156 }
4157
4158 static uint64_t
4159 em_if_get_counter(if_ctx_t ctx, ift_counter cnt)
4160 {
4161         struct e1000_softc *sc = iflib_get_softc(ctx);
4162         struct ifnet *ifp = iflib_get_ifp(ctx);
4163
4164         switch (cnt) {
4165         case IFCOUNTER_COLLISIONS:
4166                 return (sc->stats.colc);
4167         case IFCOUNTER_IERRORS:
4168                 return (sc->dropped_pkts + sc->stats.rxerrc +
4169                     sc->stats.crcerrs + sc->stats.algnerrc +
4170                     sc->stats.ruc + sc->stats.roc +
4171                     sc->stats.mpc + sc->stats.cexterr);
4172         case IFCOUNTER_OERRORS:
4173                 return (sc->stats.ecol + sc->stats.latecol +
4174                     sc->watchdog_events);
4175         default:
4176                 return (if_get_counter_default(ifp, cnt));
4177         }
4178 }
4179
4180 /* em_if_needs_restart - Tell iflib when the driver needs to be reinitialized
4181  * @ctx: iflib context
4182  * @event: event code to check
4183  *
4184  * Defaults to returning true for unknown events.
4185  *
4186  * @returns true if iflib needs to reinit the interface
4187  */
4188 static bool
4189 em_if_needs_restart(if_ctx_t ctx __unused, enum iflib_restart_event event)
4190 {
4191         switch (event) {
4192         case IFLIB_RESTART_VLAN_CONFIG:
4193                 return (false);
4194         default:
4195                 return (true);
4196         }
4197 }
4198
4199 /* Export a single 32-bit register via a read-only sysctl. */
4200 static int
4201 em_sysctl_reg_handler(SYSCTL_HANDLER_ARGS)
4202 {
4203         struct e1000_softc *sc;
4204         u_int val;
4205
4206         sc = oidp->oid_arg1;
4207         val = E1000_READ_REG(&sc->hw, oidp->oid_arg2);
4208         return (sysctl_handle_int(oidp, &val, 0, req));
4209 }
4210
4211 /*
4212  * Add sysctl variables, one per statistic, to the system.
4213  */
4214 static void
4215 em_add_hw_stats(struct e1000_softc *sc)
4216 {
4217         device_t dev = iflib_get_dev(sc->ctx);
4218         struct em_tx_queue *tx_que = sc->tx_queues;
4219         struct em_rx_queue *rx_que = sc->rx_queues;
4220
4221         struct sysctl_ctx_list *ctx = device_get_sysctl_ctx(dev);
4222         struct sysctl_oid *tree = device_get_sysctl_tree(dev);
4223         struct sysctl_oid_list *child = SYSCTL_CHILDREN(tree);
4224         struct e1000_hw_stats *stats = &sc->stats;
4225
4226         struct sysctl_oid *stat_node, *queue_node, *int_node;
4227         struct sysctl_oid_list *stat_list, *queue_list, *int_list;
4228
4229 #define QUEUE_NAME_LEN 32
4230         char namebuf[QUEUE_NAME_LEN];
4231
4232         /* Driver Statistics */
4233         SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "dropped",
4234                         CTLFLAG_RD, &sc->dropped_pkts,
4235                         "Driver dropped packets");
4236         SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "link_irq",
4237                         CTLFLAG_RD, &sc->link_irq,
4238                         "Link MSI-X IRQ Handled");
4239         SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "rx_overruns",
4240                         CTLFLAG_RD, &sc->rx_overruns,
4241                         "RX overruns");
4242         SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "watchdog_timeouts",
4243                         CTLFLAG_RD, &sc->watchdog_events,
4244                         "Watchdog timeouts");
4245         SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "device_control",
4246             CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT,
4247             sc, E1000_CTRL, em_sysctl_reg_handler, "IU",
4248             "Device Control Register");
4249         SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "rx_control",
4250             CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT,
4251             sc, E1000_RCTL, em_sysctl_reg_handler, "IU",
4252             "Receiver Control Register");
4253         SYSCTL_ADD_UINT(ctx, child, OID_AUTO, "fc_high_water",
4254                         CTLFLAG_RD, &sc->hw.fc.high_water, 0,
4255                         "Flow Control High Watermark");
4256         SYSCTL_ADD_UINT(ctx, child, OID_AUTO, "fc_low_water",
4257                         CTLFLAG_RD, &sc->hw.fc.low_water, 0,
4258                         "Flow Control Low Watermark");
4259
4260         for (int i = 0; i < sc->tx_num_queues; i++, tx_que++) {
4261                 struct tx_ring *txr = &tx_que->txr;
4262                 snprintf(namebuf, QUEUE_NAME_LEN, "queue_tx_%d", i);
4263                 queue_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, namebuf,
4264                     CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "TX Queue Name");
4265                 queue_list = SYSCTL_CHILDREN(queue_node);
4266
4267                 SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "txd_head",
4268                     CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT, sc,
4269                     E1000_TDH(txr->me), em_sysctl_reg_handler, "IU",
4270                     "Transmit Descriptor Head");
4271                 SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "txd_tail",
4272                     CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT, sc,
4273                     E1000_TDT(txr->me), em_sysctl_reg_handler, "IU",
4274                     "Transmit Descriptor Tail");
4275                 SYSCTL_ADD_ULONG(ctx, queue_list, OID_AUTO, "tx_irq",
4276                                 CTLFLAG_RD, &txr->tx_irq,
4277                                 "Queue MSI-X Transmit Interrupts");
4278         }
4279
4280         for (int j = 0; j < sc->rx_num_queues; j++, rx_que++) {
4281                 struct rx_ring *rxr = &rx_que->rxr;
4282                 snprintf(namebuf, QUEUE_NAME_LEN, "queue_rx_%d", j);
4283                 queue_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, namebuf,
4284                     CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "RX Queue Name");
4285                 queue_list = SYSCTL_CHILDREN(queue_node);
4286
4287                 SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "rxd_head",
4288                     CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT, sc,
4289                     E1000_RDH(rxr->me), em_sysctl_reg_handler, "IU",
4290                     "Receive Descriptor Head");
4291                 SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "rxd_tail",
4292                     CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT, sc,
4293                     E1000_RDT(rxr->me), em_sysctl_reg_handler, "IU",
4294                     "Receive Descriptor Tail");
4295                 SYSCTL_ADD_ULONG(ctx, queue_list, OID_AUTO, "rx_irq",
4296                                 CTLFLAG_RD, &rxr->rx_irq,
4297                                 "Queue MSI-X Receive Interrupts");
4298         }
4299
4300         /* MAC stats get their own sub node */
4301
4302         stat_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, "mac_stats",
4303             CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "Statistics");
4304         stat_list = SYSCTL_CHILDREN(stat_node);
4305
4306         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "excess_coll",
4307                         CTLFLAG_RD, &stats->ecol,
4308                         "Excessive collisions");
4309         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "single_coll",
4310                         CTLFLAG_RD, &stats->scc,
4311                         "Single collisions");
4312         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "multiple_coll",
4313                         CTLFLAG_RD, &stats->mcc,
4314                         "Multiple collisions");
4315         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "late_coll",
4316                         CTLFLAG_RD, &stats->latecol,
4317                         "Late collisions");
4318         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "collision_count",
4319                         CTLFLAG_RD, &stats->colc,
4320                         "Collision Count");
4321         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "symbol_errors",
4322                         CTLFLAG_RD, &sc->stats.symerrs,
4323                         "Symbol Errors");
4324         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "sequence_errors",
4325                         CTLFLAG_RD, &sc->stats.sec,
4326                         "Sequence Errors");
4327         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "defer_count",
4328                         CTLFLAG_RD, &sc->stats.dc,
4329                         "Defer Count");
4330         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "missed_packets",
4331                         CTLFLAG_RD, &sc->stats.mpc,
4332                         "Missed Packets");
4333         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_no_buff",
4334                         CTLFLAG_RD, &sc->stats.rnbc,
4335                         "Receive No Buffers");
4336         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_undersize",
4337                         CTLFLAG_RD, &sc->stats.ruc,
4338                         "Receive Undersize");
4339         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_fragmented",
4340                         CTLFLAG_RD, &sc->stats.rfc,
4341                         "Fragmented Packets Received ");
4342         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_oversize",
4343                         CTLFLAG_RD, &sc->stats.roc,
4344                         "Oversized Packets Received");
4345         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_jabber",
4346                         CTLFLAG_RD, &sc->stats.rjc,
4347                         "Recevied Jabber");
4348         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_errs",
4349                         CTLFLAG_RD, &sc->stats.rxerrc,
4350                         "Receive Errors");
4351         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "crc_errs",
4352                         CTLFLAG_RD, &sc->stats.crcerrs,
4353                         "CRC errors");
4354         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "alignment_errs",
4355                         CTLFLAG_RD, &sc->stats.algnerrc,
4356                         "Alignment Errors");
4357         /* On 82575 these are collision counts */
4358         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "coll_ext_errs",
4359                         CTLFLAG_RD, &sc->stats.cexterr,
4360                         "Collision/Carrier extension errors");
4361         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xon_recvd",
4362                         CTLFLAG_RD, &sc->stats.xonrxc,
4363                         "XON Received");
4364         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xon_txd",
4365                         CTLFLAG_RD, &sc->stats.xontxc,
4366                         "XON Transmitted");
4367         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xoff_recvd",
4368                         CTLFLAG_RD, &sc->stats.xoffrxc,
4369                         "XOFF Received");
4370         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xoff_txd",
4371                         CTLFLAG_RD, &sc->stats.xofftxc,
4372                         "XOFF Transmitted");
4373
4374         /* Packet Reception Stats */
4375         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_pkts_recvd",
4376                         CTLFLAG_RD, &sc->stats.tpr,
4377                         "Total Packets Received ");
4378         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_pkts_recvd",
4379                         CTLFLAG_RD, &sc->stats.gprc,
4380                         "Good Packets Received");
4381         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "bcast_pkts_recvd",
4382                         CTLFLAG_RD, &sc->stats.bprc,
4383                         "Broadcast Packets Received");
4384         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "mcast_pkts_recvd",
4385                         CTLFLAG_RD, &sc->stats.mprc,
4386                         "Multicast Packets Received");
4387         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_64",
4388                         CTLFLAG_RD, &sc->stats.prc64,
4389                         "64 byte frames received ");
4390         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_65_127",
4391                         CTLFLAG_RD, &sc->stats.prc127,
4392                         "65-127 byte frames received");
4393         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_128_255",
4394                         CTLFLAG_RD, &sc->stats.prc255,
4395                         "128-255 byte frames received");
4396         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_256_511",
4397                         CTLFLAG_RD, &sc->stats.prc511,
4398                         "256-511 byte frames received");
4399         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_512_1023",
4400                         CTLFLAG_RD, &sc->stats.prc1023,
4401                         "512-1023 byte frames received");
4402         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_1024_1522",
4403                         CTLFLAG_RD, &sc->stats.prc1522,
4404                         "1023-1522 byte frames received");
4405         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_octets_recvd",
4406                         CTLFLAG_RD, &sc->stats.gorc,
4407                         "Good Octets Received");
4408
4409         /* Packet Transmission Stats */
4410         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_octets_txd",
4411                         CTLFLAG_RD, &sc->stats.gotc,
4412                         "Good Octets Transmitted");
4413         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_pkts_txd",
4414                         CTLFLAG_RD, &sc->stats.tpt,
4415                         "Total Packets Transmitted");
4416         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_pkts_txd",
4417                         CTLFLAG_RD, &sc->stats.gptc,
4418                         "Good Packets Transmitted");
4419         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "bcast_pkts_txd",
4420                         CTLFLAG_RD, &sc->stats.bptc,
4421                         "Broadcast Packets Transmitted");
4422         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "mcast_pkts_txd",
4423                         CTLFLAG_RD, &sc->stats.mptc,
4424                         "Multicast Packets Transmitted");
4425         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_64",
4426                         CTLFLAG_RD, &sc->stats.ptc64,
4427                         "64 byte frames transmitted ");
4428         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_65_127",
4429                         CTLFLAG_RD, &sc->stats.ptc127,
4430                         "65-127 byte frames transmitted");
4431         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_128_255",
4432                         CTLFLAG_RD, &sc->stats.ptc255,
4433                         "128-255 byte frames transmitted");
4434         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_256_511",
4435                         CTLFLAG_RD, &sc->stats.ptc511,
4436                         "256-511 byte frames transmitted");
4437         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_512_1023",
4438                         CTLFLAG_RD, &sc->stats.ptc1023,
4439                         "512-1023 byte frames transmitted");
4440         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_1024_1522",
4441                         CTLFLAG_RD, &sc->stats.ptc1522,
4442                         "1024-1522 byte frames transmitted");
4443         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tso_txd",
4444                         CTLFLAG_RD, &sc->stats.tsctc,
4445                         "TSO Contexts Transmitted");
4446         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tso_ctx_fail",
4447                         CTLFLAG_RD, &sc->stats.tsctfc,
4448                         "TSO Contexts Failed");
4449
4450
4451         /* Interrupt Stats */
4452
4453         int_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, "interrupts",
4454             CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "Interrupt Statistics");
4455         int_list = SYSCTL_CHILDREN(int_node);
4456
4457         SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "asserts",
4458                         CTLFLAG_RD, &sc->stats.iac,
4459                         "Interrupt Assertion Count");
4460
4461         SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_pkt_timer",
4462                         CTLFLAG_RD, &sc->stats.icrxptc,
4463                         "Interrupt Cause Rx Pkt Timer Expire Count");
4464
4465         SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_abs_timer",
4466                         CTLFLAG_RD, &sc->stats.icrxatc,
4467                         "Interrupt Cause Rx Abs Timer Expire Count");
4468
4469         SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_pkt_timer",
4470                         CTLFLAG_RD, &sc->stats.ictxptc,
4471                         "Interrupt Cause Tx Pkt Timer Expire Count");
4472
4473         SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_abs_timer",
4474                         CTLFLAG_RD, &sc->stats.ictxatc,
4475                         "Interrupt Cause Tx Abs Timer Expire Count");
4476
4477         SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_queue_empty",
4478                         CTLFLAG_RD, &sc->stats.ictxqec,
4479                         "Interrupt Cause Tx Queue Empty Count");
4480
4481         SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_queue_min_thresh",
4482                         CTLFLAG_RD, &sc->stats.ictxqmtc,
4483                         "Interrupt Cause Tx Queue Min Thresh Count");
4484
4485         SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_desc_min_thresh",
4486                         CTLFLAG_RD, &sc->stats.icrxdmtc,
4487                         "Interrupt Cause Rx Desc Min Thresh Count");
4488
4489         SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_overrun",
4490                         CTLFLAG_RD, &sc->stats.icrxoc,
4491                         "Interrupt Cause Receiver Overrun Count");
4492 }
4493
4494 static void
4495 em_fw_version_locked(if_ctx_t ctx)
4496 {
4497         struct e1000_softc *sc = iflib_get_softc(ctx);
4498         struct e1000_hw *hw = &sc->hw;
4499         struct e1000_fw_version *fw_ver = &sc->fw_ver;
4500         uint16_t eep = 0;
4501
4502         /*
4503          * em_fw_version_locked() must run under the IFLIB_CTX_LOCK to meet the
4504          * NVM locking model, so we do it in em_if_attach_pre() and store the
4505          * info in the softc
4506          */
4507         ASSERT_CTX_LOCK_HELD(hw);
4508
4509         *fw_ver = (struct e1000_fw_version){0};
4510
4511         if (hw->mac.type >= igb_mac_min) {
4512                 /*
4513                  * Use the Shared Code for igb(4)
4514                  */
4515                 e1000_get_fw_version(hw, fw_ver);
4516         } else {
4517                 /*
4518                  * Otherwise, EEPROM version should be present on (almost?) all
4519                  * devices here
4520                  */
4521                 if(e1000_read_nvm(hw, NVM_VERSION, 1, &eep)) {
4522                         INIT_DEBUGOUT("can't get EEPROM version");
4523                         return;
4524                 }
4525
4526                 fw_ver->eep_major = (eep & NVM_MAJOR_MASK) >> NVM_MAJOR_SHIFT;
4527                 fw_ver->eep_minor = (eep & NVM_MINOR_MASK) >> NVM_MINOR_SHIFT;
4528                 fw_ver->eep_build = (eep & NVM_IMAGE_ID_MASK);
4529         }
4530 }
4531
4532 static void
4533 em_sbuf_fw_version(struct e1000_fw_version *fw_ver, struct sbuf *buf)
4534 {
4535         const char *space = "";
4536
4537         if (fw_ver->eep_major || fw_ver->eep_minor || fw_ver->eep_build) {
4538                 sbuf_printf(buf, "EEPROM V%d.%d-%d", fw_ver->eep_major,
4539                             fw_ver->eep_minor, fw_ver->eep_build);
4540                 space = " ";
4541         }
4542
4543         if (fw_ver->invm_major || fw_ver->invm_minor || fw_ver->invm_img_type) {
4544                 sbuf_printf(buf, "%sNVM V%d.%d imgtype%d",
4545                             space, fw_ver->invm_major, fw_ver->invm_minor,
4546                             fw_ver->invm_img_type);
4547                 space = " ";
4548         }
4549
4550         if (fw_ver->or_valid) {
4551                 sbuf_printf(buf, "%sOption ROM V%d-b%d-p%d",
4552                             space, fw_ver->or_major, fw_ver->or_build,
4553                             fw_ver->or_patch);
4554                 space = " ";
4555         }
4556
4557         if (fw_ver->etrack_id)
4558                 sbuf_printf(buf, "%seTrack 0x%08x", space, fw_ver->etrack_id);
4559 }
4560
4561 static void
4562 em_print_fw_version(struct e1000_softc *sc )
4563 {
4564         device_t dev = sc->dev;
4565         struct sbuf *buf;
4566         int error = 0;
4567
4568         buf = sbuf_new_auto();
4569         if (!buf) {
4570                 device_printf(dev, "Could not allocate sbuf for output.\n");
4571                 return;
4572         }
4573
4574         em_sbuf_fw_version(&sc->fw_ver, buf);
4575
4576         error = sbuf_finish(buf);
4577         if (error)
4578                 device_printf(dev, "Error finishing sbuf: %d\n", error);
4579         else if (sbuf_len(buf))
4580                 device_printf(dev, "%s\n", sbuf_data(buf));
4581
4582         sbuf_delete(buf);
4583 }
4584
4585 static int
4586 em_sysctl_print_fw_version(SYSCTL_HANDLER_ARGS)
4587 {
4588         struct e1000_softc *sc = (struct e1000_softc *)arg1;
4589         device_t dev = sc->dev;
4590         struct sbuf *buf;
4591         int error = 0;
4592
4593         buf = sbuf_new_for_sysctl(NULL, NULL, 128, req);
4594         if (!buf) {
4595                 device_printf(dev, "Could not allocate sbuf for output.\n");
4596                 return (ENOMEM);
4597         }
4598
4599         em_sbuf_fw_version(&sc->fw_ver, buf);
4600
4601         error = sbuf_finish(buf);
4602         if (error)
4603                 device_printf(dev, "Error finishing sbuf: %d\n", error);
4604
4605         sbuf_delete(buf);
4606
4607         return (0);
4608 }
4609
4610 /**********************************************************************
4611  *
4612  *  This routine provides a way to dump out the adapter eeprom,
4613  *  often a useful debug/service tool. This only dumps the first
4614  *  32 words, stuff that matters is in that extent.
4615  *
4616  **********************************************************************/
4617 static int
4618 em_sysctl_nvm_info(SYSCTL_HANDLER_ARGS)
4619 {
4620         struct e1000_softc *sc = (struct e1000_softc *)arg1;
4621         int error;
4622         int result;
4623
4624         result = -1;
4625         error = sysctl_handle_int(oidp, &result, 0, req);
4626
4627         if (error || !req->newptr)
4628                 return (error);
4629
4630         /*
4631          * This value will cause a hex dump of the
4632          * first 32 16-bit words of the EEPROM to
4633          * the screen.
4634          */
4635         if (result == 1)
4636                 em_print_nvm_info(sc);
4637
4638         return (error);
4639 }
4640
4641 static void
4642 em_print_nvm_info(struct e1000_softc *sc)
4643 {
4644         struct e1000_hw *hw = &sc->hw;
4645         struct sx *iflib_ctx_lock = iflib_ctx_lock_get(sc->ctx);
4646         u16 eeprom_data;
4647         int i, j, row = 0;
4648
4649         /* Its a bit crude, but it gets the job done */
4650         printf("\nInterface EEPROM Dump:\n");
4651         printf("Offset\n0x0000  ");
4652
4653         /* We rely on the IFLIB_CTX_LOCK as part of NVM locking model */
4654         sx_xlock(iflib_ctx_lock);
4655         ASSERT_CTX_LOCK_HELD(hw);
4656         for (i = 0, j = 0; i < 32; i++, j++) {
4657                 if (j == 8) { /* Make the offset block */
4658                         j = 0; ++row;
4659                         printf("\n0x00%x0  ",row);
4660                 }
4661                 e1000_read_nvm(hw, i, 1, &eeprom_data);
4662                 printf("%04x ", eeprom_data);
4663         }
4664         sx_xunlock(iflib_ctx_lock);
4665         printf("\n");
4666 }
4667
4668 static int
4669 em_sysctl_int_delay(SYSCTL_HANDLER_ARGS)
4670 {
4671         struct em_int_delay_info *info;
4672         struct e1000_softc *sc;
4673         u32 regval;
4674         int error, usecs, ticks;
4675
4676         info = (struct em_int_delay_info *) arg1;
4677         usecs = info->value;
4678         error = sysctl_handle_int(oidp, &usecs, 0, req);
4679         if (error != 0 || req->newptr == NULL)
4680                 return (error);
4681         if (usecs < 0 || usecs > EM_TICKS_TO_USECS(65535))
4682                 return (EINVAL);
4683         info->value = usecs;
4684         ticks = EM_USECS_TO_TICKS(usecs);
4685         if (info->offset == E1000_ITR)  /* units are 256ns here */
4686                 ticks *= 4;
4687
4688         sc = info->sc;
4689
4690         regval = E1000_READ_OFFSET(&sc->hw, info->offset);
4691         regval = (regval & ~0xffff) | (ticks & 0xffff);
4692         /* Handle a few special cases. */
4693         switch (info->offset) {
4694         case E1000_RDTR:
4695                 break;
4696         case E1000_TIDV:
4697                 if (ticks == 0) {
4698                         sc->txd_cmd &= ~E1000_TXD_CMD_IDE;
4699                         /* Don't write 0 into the TIDV register. */
4700                         regval++;
4701                 } else
4702                         sc->txd_cmd |= E1000_TXD_CMD_IDE;
4703                 break;
4704         }
4705         E1000_WRITE_OFFSET(&sc->hw, info->offset, regval);
4706         return (0);
4707 }
4708
4709 static void
4710 em_add_int_delay_sysctl(struct e1000_softc *sc, const char *name,
4711         const char *description, struct em_int_delay_info *info,
4712         int offset, int value)
4713 {
4714         info->sc = sc;
4715         info->offset = offset;
4716         info->value = value;
4717         SYSCTL_ADD_PROC(device_get_sysctl_ctx(sc->dev),
4718             SYSCTL_CHILDREN(device_get_sysctl_tree(sc->dev)),
4719             OID_AUTO, name, CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT,
4720             info, 0, em_sysctl_int_delay, "I", description);
4721 }
4722
4723 /*
4724  * Set flow control using sysctl:
4725  * Flow control values:
4726  *      0 - off
4727  *      1 - rx pause
4728  *      2 - tx pause
4729  *      3 - full
4730  */
4731 static int
4732 em_set_flowcntl(SYSCTL_HANDLER_ARGS)
4733 {
4734         int error;
4735         static int input = 3; /* default is full */
4736         struct e1000_softc      *sc = (struct e1000_softc *) arg1;
4737
4738         error = sysctl_handle_int(oidp, &input, 0, req);
4739
4740         if ((error) || (req->newptr == NULL))
4741                 return (error);
4742
4743         if (input == sc->fc) /* no change? */
4744                 return (error);
4745
4746         switch (input) {
4747         case e1000_fc_rx_pause:
4748         case e1000_fc_tx_pause:
4749         case e1000_fc_full:
4750         case e1000_fc_none:
4751                 sc->hw.fc.requested_mode = input;
4752                 sc->fc = input;
4753                 break;
4754         default:
4755                 /* Do nothing */
4756                 return (error);
4757         }
4758
4759         sc->hw.fc.current_mode = sc->hw.fc.requested_mode;
4760         e1000_force_mac_fc(&sc->hw);
4761         return (error);
4762 }
4763
4764 /*
4765  * Manage Energy Efficient Ethernet:
4766  * Control values:
4767  *     0/1 - enabled/disabled
4768  */
4769 static int
4770 em_sysctl_eee(SYSCTL_HANDLER_ARGS)
4771 {
4772         struct e1000_softc *sc = (struct e1000_softc *) arg1;
4773         int error, value;
4774
4775         value = sc->hw.dev_spec.ich8lan.eee_disable;
4776         error = sysctl_handle_int(oidp, &value, 0, req);
4777         if (error || req->newptr == NULL)
4778                 return (error);
4779         sc->hw.dev_spec.ich8lan.eee_disable = (value != 0);
4780         em_if_init(sc->ctx);
4781
4782         return (0);
4783 }
4784
4785 static int
4786 em_sysctl_debug_info(SYSCTL_HANDLER_ARGS)
4787 {
4788         struct e1000_softc *sc;
4789         int error;
4790         int result;
4791
4792         result = -1;
4793         error = sysctl_handle_int(oidp, &result, 0, req);
4794
4795         if (error || !req->newptr)
4796                 return (error);
4797
4798         if (result == 1) {
4799                 sc = (struct e1000_softc *) arg1;
4800                 em_print_debug_info(sc);
4801         }
4802
4803         return (error);
4804 }
4805
4806 static int
4807 em_get_rs(SYSCTL_HANDLER_ARGS)
4808 {
4809         struct e1000_softc *sc = (struct e1000_softc *) arg1;
4810         int error;
4811         int result;
4812
4813         result = 0;
4814         error = sysctl_handle_int(oidp, &result, 0, req);
4815
4816         if (error || !req->newptr || result != 1)
4817                 return (error);
4818         em_dump_rs(sc);
4819
4820         return (error);
4821 }
4822
4823 static void
4824 em_if_debug(if_ctx_t ctx)
4825 {
4826         em_dump_rs(iflib_get_softc(ctx));
4827 }
4828
4829 /*
4830  * This routine is meant to be fluid, add whatever is
4831  * needed for debugging a problem.  -jfv
4832  */
4833 static void
4834 em_print_debug_info(struct e1000_softc *sc)
4835 {
4836         device_t dev = iflib_get_dev(sc->ctx);
4837         struct ifnet *ifp = iflib_get_ifp(sc->ctx);
4838         struct tx_ring *txr = &sc->tx_queues->txr;
4839         struct rx_ring *rxr = &sc->rx_queues->rxr;
4840
4841         if (if_getdrvflags(ifp) & IFF_DRV_RUNNING)
4842                 printf("Interface is RUNNING ");
4843         else
4844                 printf("Interface is NOT RUNNING\n");
4845
4846         if (if_getdrvflags(ifp) & IFF_DRV_OACTIVE)
4847                 printf("and INACTIVE\n");
4848         else
4849                 printf("and ACTIVE\n");
4850
4851         for (int i = 0; i < sc->tx_num_queues; i++, txr++) {
4852                 device_printf(dev, "TX Queue %d ------\n", i);
4853                 device_printf(dev, "hw tdh = %d, hw tdt = %d\n",
4854                         E1000_READ_REG(&sc->hw, E1000_TDH(i)),
4855                         E1000_READ_REG(&sc->hw, E1000_TDT(i)));
4856
4857         }
4858         for (int j=0; j < sc->rx_num_queues; j++, rxr++) {
4859                 device_printf(dev, "RX Queue %d ------\n", j);
4860                 device_printf(dev, "hw rdh = %d, hw rdt = %d\n",
4861                         E1000_READ_REG(&sc->hw, E1000_RDH(j)),
4862                         E1000_READ_REG(&sc->hw, E1000_RDT(j)));
4863         }
4864 }
4865
4866 /*
4867  * 82574 only:
4868  * Write a new value to the EEPROM increasing the number of MSI-X
4869  * vectors from 3 to 5, for proper multiqueue support.
4870  */
4871 static void
4872 em_enable_vectors_82574(if_ctx_t ctx)
4873 {
4874         struct e1000_softc *sc = iflib_get_softc(ctx);
4875         struct e1000_hw *hw = &sc->hw;
4876         device_t dev = iflib_get_dev(ctx);
4877         u16 edata;
4878
4879         e1000_read_nvm(hw, EM_NVM_PCIE_CTRL, 1, &edata);
4880         if (bootverbose)
4881                 device_printf(dev, "EM_NVM_PCIE_CTRL = %#06x\n", edata);
4882         if (((edata & EM_NVM_MSIX_N_MASK) >> EM_NVM_MSIX_N_SHIFT) != 4) {
4883                 device_printf(dev, "Writing to eeprom: increasing "
4884                     "reported MSI-X vectors from 3 to 5...\n");
4885                 edata &= ~(EM_NVM_MSIX_N_MASK);
4886                 edata |= 4 << EM_NVM_MSIX_N_SHIFT;
4887                 e1000_write_nvm(hw, EM_NVM_PCIE_CTRL, 1, &edata);
4888                 e1000_update_nvm_checksum(hw);
4889                 device_printf(dev, "Writing to eeprom: done\n");
4890         }
4891 }