-
Notifications
You must be signed in to change notification settings - Fork 0
/
asm-x86_64.h
303 lines (292 loc) · 12 KB
/
asm-x86_64.h
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
// Assembly language support for x86_64 CPU.
// Bruno Haible 2016-12-28
// Copyright (C) 1997-2021 Bruno Haible <[email protected]>
//
// This program is free software: you can redistribute it and/or modify
// it under the terms of the GNU General Public License as published by
// the Free Software Foundation; either version 2 of the License, or
// (at your option) any later version.
//
// This program is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
// GNU General Public License for more details.
//
// You should have received a copy of the GNU General Public License
// along with this program. If not, see <https://www.gnu.org/licenses/>.
// In order not to have to maintain several copies of the assembly language
// code, we use some macros which expand into the correct syntax.
// These macros are:
// C(name)
// This expands to the name of the C variable or function `name'.
// On Unix BSD systems, this prepends an underscore.
// L(label)
// This expands to the name of a local label, having the name `label'.
// On Unix ELF systems, where there is no underscore, names beginning
// with an alphabetic character are automatically exported, so this
// prepends a dot. Note that when defining a label, the `:' must
// be inside the parentheses, not outside, because otherwise some
// ANSI C preprocessor inserts a space between the label and the `:',
// and some assemblers don't like this.
// R(reg)
// This expands to a reference to register `reg'. On Unix, this
// prepends a % charater.
// NUM(value)
// This expands to an immediate value. On Unix, this prepends a $
// character.
// ADDR(variable)
// This expands to an immediate value, the address of some variable
// or function. On Unix, this prepends a $ character. With MSVC,
// this prepends the keyword "OFFSET".
// ADDR_PCRELATIVE(variable)
// This expands to the address of symbol `variable', with program
// counter (%rip) relative addressing.
// About operand sizes: On Unix, a suffix to the instruction specifies the
// size of the operands (for example "movb", "movw", "movl"). With
// MSVC, there is no such suffix. Instead, the assembler infers the
// operand size from the names of the registers ("al" vs. "ax" vs.
// "eax"). This works well in most cases, but in instructions like
// "mul [esi]" the assembler guesses the operand size: "byte" by
// default. So it is better to explicitly specify the operand size
// of memory operands (prefix X1, X2, X4, X8).
// (Side note about Unix assemblers: Some Unix assemblers allow you
// to write "testb %eax,%eax" but silently treat this as
// "testb %al,%al".)
// X1
// This prefixes a memory reference of 1 byte.
// X2
// This prefixes a memory reference of 2 bytes.
// X4
// This prefixes a memory reference of 4 bytes.
// X8
// This prefixes a memory reference of 8 bytes.
// MEM(base)
// This expands to a memory reference at address `base'.
// MEM_DISP(base,displacement)
// This expands to a memory reference at address `base+displacement'.
// MEM_INDEX(base,index)
// This expands to a memory reference at address `base+index'.
// MEM_SHINDEX(base,index,size)
// This expands to a memory reference at address
// `base+index*size', where `size' is 1, 2, 4, or 8.
// MEM_DISP_SHINDEX0(displacement,index,size)
// This expands to a memory reference at address
// `displacement+index*size', where `size' is 1, 2, 4, or 8.
// MEM_DISP_SHINDEX(base,displacement,index,size)
// This expands to a memory reference at address
// `base+displacement+index*size', where `size' is 1, 2, 4, or 8.
// MEM_PCRELATIVE(variable)
// This expands to a memory reference at symbol `variable', with
// program counter (%rip) relative addressing.
// INDIR(value)
// This expands to an implicit indirection. On Unix, this prepends
// a * character.
// INSN1(mnemonic,size_suffix,dst)
// This expands to an instruction with one operand.
// INSN2(mnemonic,size_suffix,src,dst)
// This expands to an instruction with two operands. In our notation,
// `src' comes first and `dst' second, but they are reversed when
// expanding to Intel syntax. In Intel syntax, size_suffix is omitted.
// INSN2S(mnemonic,size_suffix,src,dst)
// This expands to an instruction with two operands. In our notation,
// `src' comes first and `dst' second, but they are reversed when
// expanding to Intel syntax. In Intel syntax, size_suffix is
// preserved.
// INSN2MOVXL(mnemonic,size_suffix,src,dst)
// This expands to an instruction with two operands, of type
// movsbl/movzbl, which in some syntaxes requires a second suffix.
// INSN2MOVXQ(mnemonic,size_suffix,src,dst)
// This expands to an instruction with two operands, of type
// movsbq/movzbq, which in some syntaxes requires a second suffix.
// INSN2MOVXLQ(mnemonic,size_suffix,src,dst)
// This expands to an instruction with two operands, of type
// movslq/movzlq, which in some syntaxes requires a second suffix.
// _
// For instructions which don't have a size suffix, like jump
// instructions. Expands to nothing. Needed for MSVC, which has
// problems with empty macro arguments.
// TEXT()
// Switch to the code section.
// P2ALIGN(log,max)
// Align to 2^log bytes, but insert at most max bytes.
// GLOBL(name)
// Declare `name' to be a global symbol.
// DECLARE_FUNCTION(name)
// Declare `name' to be a global function. When assembly language
// code is compiled into a shared library, ELF linkers need to know
// which symbols are functions.
// FUNBEGIN(name)
// Start the assembly language code for the C function `name'.
// FUNEND(name,size_expression)
// End the assembly language code for the C function 'name'.
// EH_FRAME_SECTION
// The arguments to the .section statement that introduces the
// exception handler section (on ELF platforms).
#ifdef _MSC_VER
// MSVC
#define C(entrypoint) entrypoint
#define L(label) $L##label
#else
#ifdef ASM_UNDERSCORE
// Mac OS X
#define C(entrypoint) _##entrypoint
#define L(label) L##label
#else
// Linux/ELF, Solaris/ELF, Windows with GNU as
#define C(entrypoint) entrypoint
#define L(label) .L##label
#endif
#endif
// The two syntaxes:
// - ATT_SYNTAX for GNU assembler version 2.
// - INTEL_SYNTAX for MS assembler.
// Note: INTEL_SYNTAX is not the same syntax as produced by "gcc masm=intel"
// as there are syntactic differences between that syntax and the one accepted
// by the MS assembler (for MEM_DISP, INDIR, P2ALIGN, FUNBEGIN, FUNEND etc.).
#ifdef _MSC_VER
// MS assembler
#define R(r) r
#define NUM(n) n
#define ADDR(a) OFFSET a
#define ADDR_PCRELATIVE(a) OFFSET a
#define X1 BYTE PTR
#define X2 WORD PTR
#define X4 DWORD PTR
#define X8 QWORD PTR
#define MEM(base) [base]
#define MEM_DISP(base,displacement) [base+(displacement)]
#define MEM_INDEX(base,index) [base+index]
#define MEM_SHINDEX(base,index,size) [base+index*size]
#define MEM_DISP_SHINDEX0(displacement,index,size) [(displacement)+index*size]
#define MEM_DISP_SHINDEX(base,displacement,index,size) [base+(displacement)+index*size]
#define MEM_PCRELATIVE(variable) variable
#define INDIR(value)value
#define INSNCONC(mnemonic,suffix)mnemonic##suffix
#define INSN1(mnemonic,size_suffix,dst)mnemonic dst
#define INSN2(mnemonic,size_suffix,src,dst)mnemonic dst,src
#define INSN2S(mnemonic,size_suffix,src,dst)INSNCONC(mnemonic,size_suffix) dst,src
#define INSN2MOVXL(mnemonic,size_suffix,src,dst)INSNCONC(mnemonic,x) dst,src
#define INSN2MOVXQ(mnemonic,size_suffix,src,dst)INSNCONC(mnemonic,x) dst,src
#define INSN2MOVXLQ(mnemonic,size_suffix,src,dst)INSNCONC(mnemonic,xd) dst,src
#else
// GNU assembler version 2
#define R(r) %r
#define NUM(n) $ n
#define ADDR(a) $##a
#define ADDR_PCRELATIVE(a) a(%rip)
#define X1
#define X2
#define X4
#define X8
#define MEM(base)(R(base))
#define MEM_DISP(base,displacement)displacement(R(base))
#define MEM_INDEX(base,index)(R(base),R(index))
#define MEM_SHINDEX(base,index,size)(R(base),R(index),size)
#define MEM_DISP_SHINDEX0(displacement,index,size)displacement(,R(index),size)
#define MEM_DISP_SHINDEX(base,displacement,index,size)displacement(R(base),R(index),size)
#define MEM_PCRELATIVE(variable) variable(%rip)
#define INDIR(value)*value
#define INSNCONC(mnemonic,size_suffix)mnemonic##size_suffix
#define INSN1(mnemonic,size_suffix,dst)INSNCONC(mnemonic,size_suffix) dst
#define INSN2(mnemonic,size_suffix,src,dst)INSNCONC(mnemonic,size_suffix) src,dst
#define INSN2S(mnemonic,size_suffix,src,dst)INSNCONC(mnemonic,size_suffix) src,dst
#define INSN2MOVXL(mnemonic,size_suffix,src,dst)INSNCONC(mnemonic,size_suffix##l) src,dst
#define INSN2MOVXQ(mnemonic,size_suffix,src,dst)INSNCONC(mnemonic,size_suffix##q) src,dst
#define INSN2MOVXLQ(mnemonic,size_suffix,src,dst)INSNCONC(mnemonic,size_suffix##q) src,dst
#endif
#define _
#ifdef _MSC_VER
// MS assembler
#define TEXT() _TEXT SEGMENT
#else
#define TEXT() .text
#endif
#ifdef _MSC_VER
// MS assembler
// There is no equivalent for "p2align 4,,7". This comes closest:
#define P2ALIGN(log,max) ALIGN 8
#else
#if defined __sun
// Solaris
#define P2ALIGN(log,max) .align 1<<log
#else
// Mac OS X, Linux, Windows with GNU as
#define P2ALIGN(log,max) .p2align log,,max
#endif
#endif
#ifdef _MSC_VER
// MS assembler
#define GLOBL(name)
#else
#define GLOBL(name) .globl name
#endif
// Define the DECLARE_FUNCTION(name) macro.
#ifdef _MSC_VER
// MS assembler
#define DECLARE_FUNCTION(name) PUBLIC name
#else
#if defined _WIN32 || defined __CYGWIN__
// Windows with GNU as
#define DECLARE_FUNCTION(name) .def C(name); .scl 2; .type 32; .endef
#else
#ifdef ASM_UNDERSCORE
// Mac OS X
#define DECLARE_FUNCTION(name)
#else
// ELF
#define DECLARE_FUNCTION(name) .type C(name),@function
#endif
#endif
#endif
// Define the FUNBEGIN(name) and FUNEND() macros.
#ifdef _MSC_VER
// MS assembler
#define FUNBEGIN(name) name PROC
#define FUNEND(name,size_expression) name ENDP
#else
#define FUNBEGIN(name) C(name):
#if (defined _WIN32 || defined __CYGWIN__) || defined(ASM_UNDERSCORE)
// Windows with GNU as, Mac OS X
#define FUNEND(name,size_expression)
#else
// ELF
#if defined __sun
// Solaris/ELF
#define FUNEND(name,size_expression) .size C(name), . - C(name)
#else
// Linux/ELF
#define FUNEND(name,size_expression) .size C(name),size_expression
#endif
#endif
#endif
// Section of frame info for exception handlers
#if defined __APPLE__ && defined __MACH__
// Mac OS X
#define EH_FRAME_SECTION __TEXT,__eh_frame,coalesced,no_toc+strip_static_syms+live_support
#else
#if defined __sun
// Solaris/ELF
#define EH_FRAME_SECTION .eh_frame,"aL",link=.text,@unwind
#else
// The eh_frame section was usually writable (flags "aw") in older OS versions,
// but can be made read-only (flags "a") under specific conditions, depending
// on the assembler's behaviour (look for EH_TABLES_CAN_BE_READ_ONLY in
// gcc-11.1.0/gcc/defaults.h).
// To determine whether the eh_frame section is writable or read-only by
// default, compile a simple program:
// gcc -S -fno-dwarf2-cfi-asm hello.c (GCC >= 4.4)
// or
// gcc -S hello.c (GCC < 4.4)
// The result is that all modern systems (at least Ubuntu >= 16.04,
// Red Hat Enterprise Linux >= 5, Fedora >= 13, Alpine Linux >= 3.7,
// FreeBSD >= 11, DragonFly BSD >= 6, NetBSD >= 7, OpenBSD >= 6)
// use a read-only eh_frame section.
#if 1
// all modern systems
#define EH_FRAME_SECTION .eh_frame,"a",@progbits
#else
// only very old systems
#define EH_FRAME_SECTION .eh_frame,"aw",@progbits
#endif
#endif
#endif