1/*-
2 * Copyright (c) 1990 The Regents of the University of California.
3 * All rights reserved.
4 *
5 * This code is derived from software contributed to Berkeley by
6 * William Jolitz.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. Neither the name of the University nor the names of its contributors
17 * may be used to endorse or promote products derived from this software
18 * without specific prior written permission.
19 *
20 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30 * SUCH DAMAGE.
31 */
32
33/*
34 * Floating Point Data Structures and Constants
35 * W. Jolitz 1/90
36 */
37
38#ifndef _X86_FPU_H_
39#define _X86_FPU_H_
40
41/* Environment information of floating point unit. */
42struct env87 {
43 int32_t en_cw; /* control word (16bits) */
44 int32_t en_sw; /* status word (16bits) */
45 int32_t en_tw; /* tag word (16bits) */
46 int32_t en_fip; /* fp instruction pointer */
47 uint16_t en_fcs; /* fp code segment selector */
48 uint16_t en_opcode; /* opcode last executed (11 bits) */
49 int32_t en_foo; /* fp operand offset */
50 int32_t en_fos; /* fp operand segment selector */
51};
52
53/* Contents of each x87 floating point accumulator. */
54struct fpacc87 {
55 uint8_t fp_bytes[10];
56};
57
58/* Floating point context. (i386 fnsave/frstor) */
59struct save87 {
60 struct env87 sv_env; /* floating point control/status */
61 struct fpacc87 sv_ac[8]; /* accumulator contents, 0-7 */
62 uint8_t sv_pad0[4]; /* saved status word (now unused) */
63 uint8_t sv_pad[64];
64};
65
66/* Contents of each SSE extended accumulator. */
67struct xmmacc {
68 uint8_t xmm_bytes[16];
69};
70
71/* Contents of the upper 16 bytes of each AVX extended accumulator. */
72struct ymmacc {
73 uint8_t ymm_bytes[16];
74};
75
76/* Rename structs below depending on machine architecture. */
77#ifdef __i386__
78#define __envxmm32 envxmm
79#else
80#define __envxmm32 envxmm32
81#define __envxmm64 envxmm
82#endif
83
84struct __envxmm32 {
85 uint16_t en_cw; /* control word (16bits) */
86 uint16_t en_sw; /* status word (16bits) */
87 uint16_t en_tw; /* tag word (16bits) */
88 uint16_t en_opcode; /* opcode last executed (11 bits) */
89 uint32_t en_fip; /* fp instruction pointer */
90 uint16_t en_fcs; /* fp code segment selector */
91 uint16_t en_pad0; /* padding */
92 uint32_t en_foo; /* fp operand offset */
93 uint16_t en_fos; /* fp operand segment selector */
94 uint16_t en_pad1; /* padding */
95 uint32_t en_mxcsr; /* SSE control/status register */
96 uint32_t en_mxcsr_mask; /* valid bits in mxcsr */
97};
98
99struct __envxmm64 {
100 uint16_t en_cw; /* control word (16bits) */
101 uint16_t en_sw; /* status word (16bits) */
102 uint8_t en_tw; /* tag word (8bits) */
103 uint8_t en_zero;
104 uint16_t en_opcode; /* opcode last executed (11 bits ) */
105 uint64_t en_rip; /* fp instruction pointer */
106 uint64_t en_rdp; /* fp operand pointer */
107 uint32_t en_mxcsr; /* SSE control/status register */
108 uint32_t en_mxcsr_mask; /* valid bits in mxcsr */
109};
110
111/* Floating point context. (i386 fxsave/fxrstor) */
112struct savexmm {
113 struct __envxmm32 sv_env;
114 struct {
115 struct fpacc87 fp_acc;
116 uint8_t fp_pad[6]; /* padding */
117 } sv_fp[8];
118 struct xmmacc sv_xmm[8];
119 uint8_t sv_pad[224];
120} __aligned(16);
121
122#ifdef __i386__
123union savefpu {
124 struct save87 sv_87;
125 struct savexmm sv_xmm;
126};
127#else
128/* Floating point context. (amd64 fxsave/fxrstor) */
129struct savefpu {
130 struct __envxmm64 sv_env;
131 struct {
132 struct fpacc87 fp_acc;
133 uint8_t fp_pad[6]; /* padding */
134 } sv_fp[8];
135 struct xmmacc sv_xmm[16];
136 uint8_t sv_pad[96];
137} __aligned(16);
138#endif
139
140struct xstate_hdr {
141 uint64_t xstate_bv;
142 uint64_t xstate_xcomp_bv;
143 uint8_t xstate_rsrv0[8];
144 uint8_t xstate_rsrv[40];
145};
146#define XSTATE_XCOMP_BV_COMPACT (1ULL << 63)
147
148struct savexmm_xstate {
149 struct xstate_hdr sx_hd;
150 struct ymmacc sx_ymm[16];
151};
152
153struct savexmm_ymm {
154 struct __envxmm32 sv_env;
155 struct {
156 struct fpacc87 fp_acc;
157 int8_t fp_pad[6]; /* padding */
158 } sv_fp[8];
159 struct xmmacc sv_xmm[16];
160 uint8_t sv_pad[96];
161 struct savexmm_xstate sv_xstate;
162} __aligned(64);
163
164struct savefpu_xstate {
165 struct xstate_hdr sx_hd;
166 struct ymmacc sx_ymm[16];
167};
168
169struct savefpu_ymm {
170 struct __envxmm64 sv_env;
171 struct {
172 struct fpacc87 fp_acc;
173 int8_t fp_pad[6]; /* padding */
174 } sv_fp[8];
175 struct xmmacc sv_xmm[16];
176 uint8_t sv_pad[96];
177 struct savefpu_xstate sv_xstate;
178} __aligned(64);
179
180#undef __envxmm32
181#undef __envxmm64
182
183/*
184 * The hardware default control word for i387's and later coprocessors is
185 * 0x37F, giving:
186 *
187 * round to nearest
188 * 64-bit precision
189 * all exceptions masked.
190 *
191 * FreeBSD/i386 uses 53 bit precision for things like fadd/fsub/fsqrt etc
192 * because of the difference between memory and fpu register stack arguments.
193 * If its using an intermediate fpu register, it has 80/64 bits to work
194 * with. If it uses memory, it has 64/53 bits to work with. However,
195 * gcc is aware of this and goes to a fair bit of trouble to make the
196 * best use of it.
197 *
198 * This is mostly academic for AMD64, because the ABI prefers the use
199 * SSE2 based math. For FreeBSD/amd64, we go with the default settings.
200 */
201#define __INITIAL_FPUCW__ 0x037F
202#define __INITIAL_FPUCW_I386__ 0x127F
203#define __INITIAL_NPXCW__ __INITIAL_FPUCW_I386__
204#define __INITIAL_MXCSR__ 0x1F80
205#define __INITIAL_MXCSR_MASK__ 0xFFBF
206
207/*
208 * The current value of %xcr0 is saved in the sv_pad[] field of the FPU
209 * state in the NT_X86_XSTATE note in core dumps. This offset is chosen
210 * to match the offset used by NT_X86_XSTATE in other systems.
211 */
212#define X86_XSTATE_XCR0_OFFSET 464
213
214#ifdef _KERNEL
215/*
216 * CR0_MP and CR0_EM are always set. Use CR0_TS to force traps when
217 * FPU access is disabled.
218 */
219#define fpu_enable() clts()
220#define fpu_disable() load_cr0(rcr0() | CR0_TS)
221
222bool xsave_extfeature_supported(uint64_t feature, bool supervisor);
223bool xsave_extension_supported(uint64_t extension);
224size_t xsave_area_hdr_offset(void);
225size_t xsave_area_offset(uint64_t xstate_bv, uint64_t feature, bool compact,
226 bool supervisor);
227size_t xsave_area_size(uint64_t xstate_bv, bool compact, bool supervisor);
228#endif
229
230#endif /* !_X86_FPU_H_ */