• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 //===-- X86Disassembler.cpp - Disassembler for x86 and x86_64 -------------===//
2 //
3 //                     The LLVM Compiler Infrastructure
4 //
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
7 //
8 //===----------------------------------------------------------------------===//
9 //
10 // This file is part of the X86 Disassembler.
11 // It contains code to translate the data produced by the decoder into
12 //  MCInsts.
13 // Documentation for the disassembler can be found in X86Disassembler.h.
14 //
15 //===----------------------------------------------------------------------===//
16 
17 /* Capstone Disassembly Engine */
18 /* By Nguyen Anh Quynh <aquynh@gmail.com>, 2013-2015 */
19 
20 #ifdef CAPSTONE_HAS_X86
21 
22 #if defined (WIN32) || defined (WIN64) || defined (_WIN32) || defined (_WIN64)
23 #pragma warning(disable:4996)			// disable MSVC's warning on strncpy()
24 #pragma warning(disable:28719)		// disable MSVC's warning on strncpy()
25 #endif
26 
27 #include <capstone/platform.h>
28 
29 #if defined(CAPSTONE_HAS_OSXKERNEL)
30 #include <Availability.h>
31 #endif
32 
33 #include <string.h>
34 
35 #include "../../cs_priv.h"
36 
37 #include "X86Disassembler.h"
38 #include "X86DisassemblerDecoderCommon.h"
39 #include "X86DisassemblerDecoder.h"
40 #include "../../MCInst.h"
41 #include "../../utils.h"
42 #include "X86Mapping.h"
43 
44 #define GET_REGINFO_ENUM
45 #define GET_REGINFO_MC_DESC
46 #include "X86GenRegisterInfo.inc"
47 
48 #define GET_INSTRINFO_ENUM
49 #ifdef CAPSTONE_X86_REDUCE
50 #include "X86GenInstrInfo_reduce.inc"
51 #else
52 #include "X86GenInstrInfo.inc"
53 #endif
54 
55 // Fill-ins to make the compiler happy.  These constants are never actually
56 //   assigned; they are just filler to make an automatically-generated switch
57 //   statement work.
58 enum {
59 	X86_BX_SI = 500,
60 	X86_BX_DI = 501,
61 	X86_BP_SI = 502,
62 	X86_BP_DI = 503,
63 	X86_sib   = 504,
64 	X86_sib64 = 505
65 };
66 
67 //
68 // Private code that translates from struct InternalInstructions to MCInsts.
69 //
70 
71 /// translateRegister - Translates an internal register to the appropriate LLVM
72 ///   register, and appends it as an operand to an MCInst.
73 ///
74 /// @param mcInst     - The MCInst to append to.
75 /// @param reg        - The Reg to append.
translateRegister(MCInst * mcInst,Reg reg)76 static void translateRegister(MCInst *mcInst, Reg reg)
77 {
78 #define ENTRY(x) X86_##x,
79 	static const uint8_t llvmRegnums[] = {
80 		ALL_REGS
81 			0
82 	};
83 #undef ENTRY
84 
85 	uint8_t llvmRegnum = llvmRegnums[reg];
86 	MCOperand_CreateReg0(mcInst, llvmRegnum);
87 }
88 
89 static const uint8_t segmentRegnums[SEG_OVERRIDE_max] = {
90 	0,        // SEG_OVERRIDE_NONE
91 	X86_CS,
92 	X86_SS,
93 	X86_DS,
94 	X86_ES,
95 	X86_FS,
96 	X86_GS
97 };
98 
99 /// translateSrcIndex   - Appends a source index operand to an MCInst.
100 ///
101 /// @param mcInst       - The MCInst to append to.
102 /// @param insn         - The internal instruction.
translateSrcIndex(MCInst * mcInst,InternalInstruction * insn)103 static bool translateSrcIndex(MCInst *mcInst, InternalInstruction *insn)
104 {
105 	unsigned baseRegNo;
106 
107 	if (insn->mode == MODE_64BIT)
108 		baseRegNo = insn->isPrefix67 ? X86_ESI : X86_RSI;
109 	else if (insn->mode == MODE_32BIT)
110 		baseRegNo = insn->isPrefix67 ? X86_SI : X86_ESI;
111 	else {
112 		// assert(insn->mode == MODE_16BIT);
113 		baseRegNo = insn->isPrefix67 ? X86_ESI : X86_SI;
114 	}
115 
116 	MCOperand_CreateReg0(mcInst, baseRegNo);
117 
118 	MCOperand_CreateReg0(mcInst, segmentRegnums[insn->segmentOverride]);
119 
120 	return false;
121 }
122 
123 /// translateDstIndex   - Appends a destination index operand to an MCInst.
124 ///
125 /// @param mcInst       - The MCInst to append to.
126 /// @param insn         - The internal instruction.
translateDstIndex(MCInst * mcInst,InternalInstruction * insn)127 static bool translateDstIndex(MCInst *mcInst, InternalInstruction *insn)
128 {
129 	unsigned baseRegNo;
130 
131 	if (insn->mode == MODE_64BIT)
132 		baseRegNo = insn->isPrefix67 ? X86_EDI : X86_RDI;
133 	else if (insn->mode == MODE_32BIT)
134 		baseRegNo = insn->isPrefix67 ? X86_DI : X86_EDI;
135 	else {
136 		// assert(insn->mode == MODE_16BIT);
137 		baseRegNo = insn->isPrefix67 ? X86_EDI : X86_DI;
138 	}
139 
140 	MCOperand_CreateReg0(mcInst, baseRegNo);
141 
142 	return false;
143 }
144 
145 /// translateImmediate  - Appends an immediate operand to an MCInst.
146 ///
147 /// @param mcInst       - The MCInst to append to.
148 /// @param immediate    - The immediate value to append.
149 /// @param operand      - The operand, as stored in the descriptor table.
150 /// @param insn         - The internal instruction.
translateImmediate(MCInst * mcInst,uint64_t immediate,const OperandSpecifier * operand,InternalInstruction * insn)151 static void translateImmediate(MCInst *mcInst, uint64_t immediate,
152 		const OperandSpecifier *operand, InternalInstruction *insn)
153 {
154 	OperandType type;
155 
156 	type = (OperandType)operand->type;
157 	if (type == TYPE_RELv) {
158 		//isBranch = true;
159 		//pcrel = insn->startLocation + insn->immediateOffset + insn->immediateSize;
160 		switch (insn->displacementSize) {
161 			case 1:
162 				if (immediate & 0x80)
163 					immediate |= ~(0xffull);
164 				break;
165 			case 2:
166 				if (immediate & 0x8000)
167 					immediate |= ~(0xffffull);
168 				break;
169 			case 4:
170 				if (immediate & 0x80000000)
171 					immediate |= ~(0xffffffffull);
172 				break;
173 			case 8:
174 				break;
175 			default:
176 				break;
177 		}
178 	} // By default sign-extend all X86 immediates based on their encoding.
179 	else if (type == TYPE_IMM8 || type == TYPE_IMM16 || type == TYPE_IMM32 ||
180 			type == TYPE_IMM64 || type == TYPE_IMMv) {
181 
182 		switch (operand->encoding) {
183 			default:
184 				break;
185 			case ENCODING_IB:
186 				if(immediate & 0x80)
187 					immediate |= ~(0xffull);
188 				break;
189 			case ENCODING_IW:
190 				if(immediate & 0x8000)
191 					immediate |= ~(0xffffull);
192 				break;
193 			case ENCODING_ID:
194 				if(immediate & 0x80000000)
195 					immediate |= ~(0xffffffffull);
196 				break;
197 			case ENCODING_IO:
198 				break;
199 		}
200 	} else if (type == TYPE_IMM3) {
201 #ifndef CAPSTONE_X86_REDUCE
202 		// Check for immediates that printSSECC can't handle.
203 		if (immediate >= 8) {
204 			unsigned NewOpc = 0;
205 
206 			switch (MCInst_getOpcode(mcInst)) {
207 				default: break;	// never reach
208 				case X86_CMPPDrmi:  NewOpc = X86_CMPPDrmi_alt;  break;
209 				case X86_CMPPDrri:  NewOpc = X86_CMPPDrri_alt;  break;
210 				case X86_CMPPSrmi:  NewOpc = X86_CMPPSrmi_alt;  break;
211 				case X86_CMPPSrri:  NewOpc = X86_CMPPSrri_alt;  break;
212 				case X86_CMPSDrm:   NewOpc = X86_CMPSDrm_alt;   break;
213 				case X86_CMPSDrr:   NewOpc = X86_CMPSDrr_alt;   break;
214 				case X86_CMPSSrm:   NewOpc = X86_CMPSSrm_alt;   break;
215 				case X86_CMPSSrr:   NewOpc = X86_CMPSSrr_alt;   break;
216 				case X86_VPCOMBri:  NewOpc = X86_VPCOMBri_alt;  break;
217 				case X86_VPCOMBmi:  NewOpc = X86_VPCOMBmi_alt;  break;
218 				case X86_VPCOMWri:  NewOpc = X86_VPCOMWri_alt;  break;
219 				case X86_VPCOMWmi:  NewOpc = X86_VPCOMWmi_alt;  break;
220 				case X86_VPCOMDri:  NewOpc = X86_VPCOMDri_alt;  break;
221 				case X86_VPCOMDmi:  NewOpc = X86_VPCOMDmi_alt;  break;
222 				case X86_VPCOMQri:  NewOpc = X86_VPCOMQri_alt;  break;
223 				case X86_VPCOMQmi:  NewOpc = X86_VPCOMQmi_alt;  break;
224 				case X86_VPCOMUBri: NewOpc = X86_VPCOMUBri_alt; break;
225 				case X86_VPCOMUBmi: NewOpc = X86_VPCOMUBmi_alt; break;
226 				case X86_VPCOMUWri: NewOpc = X86_VPCOMUWri_alt; break;
227 				case X86_VPCOMUWmi: NewOpc = X86_VPCOMUWmi_alt; break;
228 				case X86_VPCOMUDri: NewOpc = X86_VPCOMUDri_alt; break;
229 				case X86_VPCOMUDmi: NewOpc = X86_VPCOMUDmi_alt; break;
230 				case X86_VPCOMUQri: NewOpc = X86_VPCOMUQri_alt; break;
231 				case X86_VPCOMUQmi: NewOpc = X86_VPCOMUQmi_alt; break;
232 			}
233 			// Switch opcode to the one that doesn't get special printing.
234 			if (NewOpc != 0) {
235 				MCInst_setOpcode(mcInst, NewOpc);
236 			}
237 		}
238 #endif
239 	} else if (type == TYPE_IMM5) {
240 #ifndef CAPSTONE_X86_REDUCE
241 		// Check for immediates that printAVXCC can't handle.
242 		if (immediate >= 32) {
243 			unsigned NewOpc = 0;
244 
245 			switch (MCInst_getOpcode(mcInst)) {
246 				default: break; // unexpected opcode
247 				case X86_VCMPPDrmi:   NewOpc = X86_VCMPPDrmi_alt;   break;
248 				case X86_VCMPPDrri:   NewOpc = X86_VCMPPDrri_alt;   break;
249 				case X86_VCMPPSrmi:   NewOpc = X86_VCMPPSrmi_alt;   break;
250 				case X86_VCMPPSrri:   NewOpc = X86_VCMPPSrri_alt;   break;
251 				case X86_VCMPSDrm:    NewOpc = X86_VCMPSDrm_alt;    break;
252 				case X86_VCMPSDrr:    NewOpc = X86_VCMPSDrr_alt;    break;
253 				case X86_VCMPSSrm:    NewOpc = X86_VCMPSSrm_alt;    break;
254 				case X86_VCMPSSrr:    NewOpc = X86_VCMPSSrr_alt;    break;
255 				case X86_VCMPPDYrmi:  NewOpc = X86_VCMPPDYrmi_alt;  break;
256 				case X86_VCMPPDYrri:  NewOpc = X86_VCMPPDYrri_alt;  break;
257 				case X86_VCMPPSYrmi:  NewOpc = X86_VCMPPSYrmi_alt;  break;
258 				case X86_VCMPPSYrri:  NewOpc = X86_VCMPPSYrri_alt;  break;
259 				case X86_VCMPPDZrmi:  NewOpc = X86_VCMPPDZrmi_alt;  break;
260 				case X86_VCMPPDZrri:  NewOpc = X86_VCMPPDZrri_alt;  break;
261 				case X86_VCMPPDZrrib: NewOpc = X86_VCMPPDZrrib_alt; break;
262 				case X86_VCMPPSZrmi:  NewOpc = X86_VCMPPSZrmi_alt;  break;
263 				case X86_VCMPPSZrri:  NewOpc = X86_VCMPPSZrri_alt;  break;
264 				case X86_VCMPPSZrrib: NewOpc = X86_VCMPPSZrrib_alt; break;
265 				case X86_VCMPSDZrm:   NewOpc = X86_VCMPSDZrmi_alt;  break;
266 				case X86_VCMPSDZrr:   NewOpc = X86_VCMPSDZrri_alt;  break;
267 				case X86_VCMPSSZrm:   NewOpc = X86_VCMPSSZrmi_alt;  break;
268 				case X86_VCMPSSZrr:   NewOpc = X86_VCMPSSZrri_alt;  break;
269 			}
270 			// Switch opcode to the one that doesn't get special printing.
271 			if (NewOpc != 0) {
272 				MCInst_setOpcode(mcInst, NewOpc);
273 			}
274 		}
275 #endif
276 	} else if (type == TYPE_AVX512ICC) {
277 #ifndef CAPSTONE_X86_REDUCE
278 		if (immediate >= 8 || ((immediate & 0x3) == 3)) {
279 			unsigned NewOpc = 0;
280 			switch (MCInst_getOpcode(mcInst)) {
281 				default: // llvm_unreachable("unexpected opcode");
282 				case X86_VPCMPBZ128rmi:    NewOpc = X86_VPCMPBZ128rmi_alt;    break;
283 				case X86_VPCMPBZ128rmik:   NewOpc = X86_VPCMPBZ128rmik_alt;   break;
284 				case X86_VPCMPBZ128rri:    NewOpc = X86_VPCMPBZ128rri_alt;    break;
285 				case X86_VPCMPBZ128rrik:   NewOpc = X86_VPCMPBZ128rrik_alt;   break;
286 				case X86_VPCMPBZ256rmi:    NewOpc = X86_VPCMPBZ256rmi_alt;    break;
287 				case X86_VPCMPBZ256rmik:   NewOpc = X86_VPCMPBZ256rmik_alt;   break;
288 				case X86_VPCMPBZ256rri:    NewOpc = X86_VPCMPBZ256rri_alt;    break;
289 				case X86_VPCMPBZ256rrik:   NewOpc = X86_VPCMPBZ256rrik_alt;   break;
290 				case X86_VPCMPBZrmi:       NewOpc = X86_VPCMPBZrmi_alt;       break;
291 				case X86_VPCMPBZrmik:      NewOpc = X86_VPCMPBZrmik_alt;      break;
292 				case X86_VPCMPBZrri:       NewOpc = X86_VPCMPBZrri_alt;       break;
293 				case X86_VPCMPBZrrik:      NewOpc = X86_VPCMPBZrrik_alt;      break;
294 				case X86_VPCMPDZ128rmi:    NewOpc = X86_VPCMPDZ128rmi_alt;    break;
295 				case X86_VPCMPDZ128rmib:   NewOpc = X86_VPCMPDZ128rmib_alt;   break;
296 				case X86_VPCMPDZ128rmibk:  NewOpc = X86_VPCMPDZ128rmibk_alt;  break;
297 				case X86_VPCMPDZ128rmik:   NewOpc = X86_VPCMPDZ128rmik_alt;   break;
298 				case X86_VPCMPDZ128rri:    NewOpc = X86_VPCMPDZ128rri_alt;    break;
299 				case X86_VPCMPDZ128rrik:   NewOpc = X86_VPCMPDZ128rrik_alt;   break;
300 				case X86_VPCMPDZ256rmi:    NewOpc = X86_VPCMPDZ256rmi_alt;    break;
301 				case X86_VPCMPDZ256rmib:   NewOpc = X86_VPCMPDZ256rmib_alt;   break;
302 				case X86_VPCMPDZ256rmibk:  NewOpc = X86_VPCMPDZ256rmibk_alt;  break;
303 				case X86_VPCMPDZ256rmik:   NewOpc = X86_VPCMPDZ256rmik_alt;   break;
304 				case X86_VPCMPDZ256rri:    NewOpc = X86_VPCMPDZ256rri_alt;    break;
305 				case X86_VPCMPDZ256rrik:   NewOpc = X86_VPCMPDZ256rrik_alt;   break;
306 				case X86_VPCMPDZrmi:       NewOpc = X86_VPCMPDZrmi_alt;       break;
307 				case X86_VPCMPDZrmib:      NewOpc = X86_VPCMPDZrmib_alt;      break;
308 				case X86_VPCMPDZrmibk:     NewOpc = X86_VPCMPDZrmibk_alt;     break;
309 				case X86_VPCMPDZrmik:      NewOpc = X86_VPCMPDZrmik_alt;      break;
310 				case X86_VPCMPDZrri:       NewOpc = X86_VPCMPDZrri_alt;       break;
311 				case X86_VPCMPDZrrik:      NewOpc = X86_VPCMPDZrrik_alt;      break;
312 				case X86_VPCMPQZ128rmi:    NewOpc = X86_VPCMPQZ128rmi_alt;    break;
313 				case X86_VPCMPQZ128rmib:   NewOpc = X86_VPCMPQZ128rmib_alt;   break;
314 				case X86_VPCMPQZ128rmibk:  NewOpc = X86_VPCMPQZ128rmibk_alt;  break;
315 				case X86_VPCMPQZ128rmik:   NewOpc = X86_VPCMPQZ128rmik_alt;   break;
316 				case X86_VPCMPQZ128rri:    NewOpc = X86_VPCMPQZ128rri_alt;    break;
317 				case X86_VPCMPQZ128rrik:   NewOpc = X86_VPCMPQZ128rrik_alt;   break;
318 				case X86_VPCMPQZ256rmi:    NewOpc = X86_VPCMPQZ256rmi_alt;    break;
319 				case X86_VPCMPQZ256rmib:   NewOpc = X86_VPCMPQZ256rmib_alt;   break;
320 				case X86_VPCMPQZ256rmibk:  NewOpc = X86_VPCMPQZ256rmibk_alt;  break;
321 				case X86_VPCMPQZ256rmik:   NewOpc = X86_VPCMPQZ256rmik_alt;   break;
322 				case X86_VPCMPQZ256rri:    NewOpc = X86_VPCMPQZ256rri_alt;    break;
323 				case X86_VPCMPQZ256rrik:   NewOpc = X86_VPCMPQZ256rrik_alt;   break;
324 				case X86_VPCMPQZrmi:       NewOpc = X86_VPCMPQZrmi_alt;       break;
325 				case X86_VPCMPQZrmib:      NewOpc = X86_VPCMPQZrmib_alt;      break;
326 				case X86_VPCMPQZrmibk:     NewOpc = X86_VPCMPQZrmibk_alt;     break;
327 				case X86_VPCMPQZrmik:      NewOpc = X86_VPCMPQZrmik_alt;      break;
328 				case X86_VPCMPQZrri:       NewOpc = X86_VPCMPQZrri_alt;       break;
329 				case X86_VPCMPQZrrik:      NewOpc = X86_VPCMPQZrrik_alt;      break;
330 				case X86_VPCMPUBZ128rmi:   NewOpc = X86_VPCMPUBZ128rmi_alt;   break;
331 				case X86_VPCMPUBZ128rmik:  NewOpc = X86_VPCMPUBZ128rmik_alt;  break;
332 				case X86_VPCMPUBZ128rri:   NewOpc = X86_VPCMPUBZ128rri_alt;   break;
333 				case X86_VPCMPUBZ128rrik:  NewOpc = X86_VPCMPUBZ128rrik_alt;  break;
334 				case X86_VPCMPUBZ256rmi:   NewOpc = X86_VPCMPUBZ256rmi_alt;   break;
335 				case X86_VPCMPUBZ256rmik:  NewOpc = X86_VPCMPUBZ256rmik_alt;  break;
336 				case X86_VPCMPUBZ256rri:   NewOpc = X86_VPCMPUBZ256rri_alt;   break;
337 				case X86_VPCMPUBZ256rrik:  NewOpc = X86_VPCMPUBZ256rrik_alt;  break;
338 				case X86_VPCMPUBZrmi:      NewOpc = X86_VPCMPUBZrmi_alt;      break;
339 				case X86_VPCMPUBZrmik:     NewOpc = X86_VPCMPUBZrmik_alt;     break;
340 				case X86_VPCMPUBZrri:      NewOpc = X86_VPCMPUBZrri_alt;      break;
341 				case X86_VPCMPUBZrrik:     NewOpc = X86_VPCMPUBZrrik_alt;     break;
342 				case X86_VPCMPUDZ128rmi:   NewOpc = X86_VPCMPUDZ128rmi_alt;   break;
343 				case X86_VPCMPUDZ128rmib:  NewOpc = X86_VPCMPUDZ128rmib_alt;  break;
344 				case X86_VPCMPUDZ128rmibk: NewOpc = X86_VPCMPUDZ128rmibk_alt; break;
345 				case X86_VPCMPUDZ128rmik:  NewOpc = X86_VPCMPUDZ128rmik_alt;  break;
346 				case X86_VPCMPUDZ128rri:   NewOpc = X86_VPCMPUDZ128rri_alt;   break;
347 				case X86_VPCMPUDZ128rrik:  NewOpc = X86_VPCMPUDZ128rrik_alt;  break;
348 				case X86_VPCMPUDZ256rmi:   NewOpc = X86_VPCMPUDZ256rmi_alt;   break;
349 				case X86_VPCMPUDZ256rmib:  NewOpc = X86_VPCMPUDZ256rmib_alt;  break;
350 				case X86_VPCMPUDZ256rmibk: NewOpc = X86_VPCMPUDZ256rmibk_alt; break;
351 				case X86_VPCMPUDZ256rmik:  NewOpc = X86_VPCMPUDZ256rmik_alt;  break;
352 				case X86_VPCMPUDZ256rri:   NewOpc = X86_VPCMPUDZ256rri_alt;   break;
353 				case X86_VPCMPUDZ256rrik:  NewOpc = X86_VPCMPUDZ256rrik_alt;  break;
354 				case X86_VPCMPUDZrmi:      NewOpc = X86_VPCMPUDZrmi_alt;      break;
355 				case X86_VPCMPUDZrmib:     NewOpc = X86_VPCMPUDZrmib_alt;     break;
356 				case X86_VPCMPUDZrmibk:    NewOpc = X86_VPCMPUDZrmibk_alt;    break;
357 				case X86_VPCMPUDZrmik:     NewOpc = X86_VPCMPUDZrmik_alt;     break;
358 				case X86_VPCMPUDZrri:      NewOpc = X86_VPCMPUDZrri_alt;      break;
359 				case X86_VPCMPUDZrrik:     NewOpc = X86_VPCMPUDZrrik_alt;     break;
360 				case X86_VPCMPUQZ128rmi:   NewOpc = X86_VPCMPUQZ128rmi_alt;   break;
361 				case X86_VPCMPUQZ128rmib:  NewOpc = X86_VPCMPUQZ128rmib_alt;  break;
362 				case X86_VPCMPUQZ128rmibk: NewOpc = X86_VPCMPUQZ128rmibk_alt; break;
363 				case X86_VPCMPUQZ128rmik:  NewOpc = X86_VPCMPUQZ128rmik_alt;  break;
364 				case X86_VPCMPUQZ128rri:   NewOpc = X86_VPCMPUQZ128rri_alt;   break;
365 				case X86_VPCMPUQZ128rrik:  NewOpc = X86_VPCMPUQZ128rrik_alt;  break;
366 				case X86_VPCMPUQZ256rmi:   NewOpc = X86_VPCMPUQZ256rmi_alt;   break;
367 				case X86_VPCMPUQZ256rmib:  NewOpc = X86_VPCMPUQZ256rmib_alt;  break;
368 				case X86_VPCMPUQZ256rmibk: NewOpc = X86_VPCMPUQZ256rmibk_alt; break;
369 				case X86_VPCMPUQZ256rmik:  NewOpc = X86_VPCMPUQZ256rmik_alt;  break;
370 				case X86_VPCMPUQZ256rri:   NewOpc = X86_VPCMPUQZ256rri_alt;   break;
371 				case X86_VPCMPUQZ256rrik:  NewOpc = X86_VPCMPUQZ256rrik_alt;  break;
372 				case X86_VPCMPUQZrmi:      NewOpc = X86_VPCMPUQZrmi_alt;      break;
373 				case X86_VPCMPUQZrmib:     NewOpc = X86_VPCMPUQZrmib_alt;     break;
374 				case X86_VPCMPUQZrmibk:    NewOpc = X86_VPCMPUQZrmibk_alt;    break;
375 				case X86_VPCMPUQZrmik:     NewOpc = X86_VPCMPUQZrmik_alt;     break;
376 				case X86_VPCMPUQZrri:      NewOpc = X86_VPCMPUQZrri_alt;      break;
377 				case X86_VPCMPUQZrrik:     NewOpc = X86_VPCMPUQZrrik_alt;     break;
378 				case X86_VPCMPUWZ128rmi:   NewOpc = X86_VPCMPUWZ128rmi_alt;   break;
379 				case X86_VPCMPUWZ128rmik:  NewOpc = X86_VPCMPUWZ128rmik_alt;  break;
380 				case X86_VPCMPUWZ128rri:   NewOpc = X86_VPCMPUWZ128rri_alt;   break;
381 				case X86_VPCMPUWZ128rrik:  NewOpc = X86_VPCMPUWZ128rrik_alt;  break;
382 				case X86_VPCMPUWZ256rmi:   NewOpc = X86_VPCMPUWZ256rmi_alt;   break;
383 				case X86_VPCMPUWZ256rmik:  NewOpc = X86_VPCMPUWZ256rmik_alt;  break;
384 				case X86_VPCMPUWZ256rri:   NewOpc = X86_VPCMPUWZ256rri_alt;   break;
385 				case X86_VPCMPUWZ256rrik:  NewOpc = X86_VPCMPUWZ256rrik_alt;  break;
386 				case X86_VPCMPUWZrmi:      NewOpc = X86_VPCMPUWZrmi_alt;      break;
387 				case X86_VPCMPUWZrmik:     NewOpc = X86_VPCMPUWZrmik_alt;     break;
388 				case X86_VPCMPUWZrri:      NewOpc = X86_VPCMPUWZrri_alt;      break;
389 				case X86_VPCMPUWZrrik:     NewOpc = X86_VPCMPUWZrrik_alt;     break;
390 				case X86_VPCMPWZ128rmi:    NewOpc = X86_VPCMPWZ128rmi_alt;    break;
391 				case X86_VPCMPWZ128rmik:   NewOpc = X86_VPCMPWZ128rmik_alt;   break;
392 				case X86_VPCMPWZ128rri:    NewOpc = X86_VPCMPWZ128rri_alt;    break;
393 				case X86_VPCMPWZ128rrik:   NewOpc = X86_VPCMPWZ128rrik_alt;   break;
394 				case X86_VPCMPWZ256rmi:    NewOpc = X86_VPCMPWZ256rmi_alt;    break;
395 				case X86_VPCMPWZ256rmik:   NewOpc = X86_VPCMPWZ256rmik_alt;   break;
396 				case X86_VPCMPWZ256rri:    NewOpc = X86_VPCMPWZ256rri_alt;    break;
397 				case X86_VPCMPWZ256rrik:   NewOpc = X86_VPCMPWZ256rrik_alt;   break;
398 				case X86_VPCMPWZrmi:       NewOpc = X86_VPCMPWZrmi_alt;       break;
399 				case X86_VPCMPWZrmik:      NewOpc = X86_VPCMPWZrmik_alt;      break;
400 				case X86_VPCMPWZrri:       NewOpc = X86_VPCMPWZrri_alt;       break;
401 				case X86_VPCMPWZrrik:      NewOpc = X86_VPCMPWZrrik_alt;      break;
402 			}
403 			// Switch opcode to the one that doesn't get special printing.
404 			if (NewOpc != 0) {
405 				MCInst_setOpcode(mcInst, NewOpc);
406 			}
407 		}
408 #endif
409 	}
410 
411 	switch (type) {
412 		case TYPE_XMM32:
413 		case TYPE_XMM64:
414 		case TYPE_XMM128:
415 			MCOperand_CreateReg0(mcInst, X86_XMM0 + ((uint32_t)immediate >> 4));
416 			return;
417 		case TYPE_XMM256:
418 			MCOperand_CreateReg0(mcInst, X86_YMM0 + ((uint32_t)immediate >> 4));
419 			return;
420 		case TYPE_XMM512:
421 			MCOperand_CreateReg0(mcInst, X86_ZMM0 + ((uint32_t)immediate >> 4));
422 			return;
423 		case TYPE_REL8:
424 			if(immediate & 0x80)
425 				immediate |= ~(0xffull);
426 			break;
427 		case TYPE_REL32:
428 		case TYPE_REL64:
429 			if(immediate & 0x80000000)
430 				immediate |= ~(0xffffffffull);
431 			break;
432 		default:
433 			// operand is 64 bits wide.  Do nothing.
434 			break;
435 	}
436 
437 	MCOperand_CreateImm0(mcInst, immediate);
438 
439 	if (type == TYPE_MOFFS8 || type == TYPE_MOFFS16 ||
440 			type == TYPE_MOFFS32 || type == TYPE_MOFFS64) {
441 		MCOperand_CreateReg0(mcInst, segmentRegnums[insn->segmentOverride]);
442 	}
443 }
444 
445 /// translateRMRegister - Translates a register stored in the R/M field of the
446 ///   ModR/M byte to its LLVM equivalent and appends it to an MCInst.
447 /// @param mcInst       - The MCInst to append to.
448 /// @param insn         - The internal instruction to extract the R/M field
449 ///                       from.
450 /// @return             - 0 on success; -1 otherwise
translateRMRegister(MCInst * mcInst,InternalInstruction * insn)451 static bool translateRMRegister(MCInst *mcInst, InternalInstruction *insn)
452 {
453 	if (insn->eaBase == EA_BASE_sib || insn->eaBase == EA_BASE_sib64) {
454 		//debug("A R/M register operand may not have a SIB byte");
455 		return true;
456 	}
457 
458 	switch (insn->eaBase) {
459 		case EA_BASE_NONE:
460 			//debug("EA_BASE_NONE for ModR/M base");
461 			return true;
462 #define ENTRY(x) case EA_BASE_##x:
463 			ALL_EA_BASES
464 #undef ENTRY
465 				//debug("A R/M register operand may not have a base; "
466 				//      "the operand must be a register.");
467 				return true;
468 #define ENTRY(x)                                                      \
469 		case EA_REG_##x:                                                    \
470 			MCOperand_CreateReg0(mcInst, X86_##x); break;
471 			ALL_REGS
472 #undef ENTRY
473 		default:
474 				//debug("Unexpected EA base register");
475 				return true;
476 	}
477 
478 	return false;
479 }
480 
481 /// translateRMMemory - Translates a memory operand stored in the Mod and R/M
482 ///   fields of an internal instruction (and possibly its SIB byte) to a memory
483 ///   operand in LLVM's format, and appends it to an MCInst.
484 ///
485 /// @param mcInst       - The MCInst to append to.
486 /// @param insn         - The instruction to extract Mod, R/M, and SIB fields
487 ///                       from.
488 /// @return             - 0 on success; nonzero otherwise
translateRMMemory(MCInst * mcInst,InternalInstruction * insn)489 static bool translateRMMemory(MCInst *mcInst, InternalInstruction *insn)
490 {
491 	// Addresses in an MCInst are represented as five operands:
492 	//   1. basereg       (register)  The R/M base, or (if there is a SIB) the
493 	//                                SIB base
494 	//   2. scaleamount   (immediate) 1, or (if there is a SIB) the specified
495 	//                                scale amount
496 	//   3. indexreg      (register)  x86_registerNONE, or (if there is a SIB)
497 	//                                the index (which is multiplied by the
498 	//                                scale amount)
499 	//   4. displacement  (immediate) 0, or the displacement if there is one
500 	//   5. segmentreg    (register)  x86_registerNONE for now, but could be set
501 	//                                if we have segment overrides
502 
503 	bool IndexIs512, IndexIs128, IndexIs256;
504 	int scaleAmount, indexReg;
505 #ifndef CAPSTONE_X86_REDUCE
506 	uint32_t Opcode;
507 #endif
508 
509 	if (insn->eaBase == EA_BASE_sib || insn->eaBase == EA_BASE_sib64) {
510 		if (insn->sibBase != SIB_BASE_NONE) {
511 			switch (insn->sibBase) {
512 #define ENTRY(x)                                          \
513 				case SIB_BASE_##x:                                  \
514 				MCOperand_CreateReg0(mcInst, X86_##x); break;
515 				ALL_SIB_BASES
516 #undef ENTRY
517 				default:
518 					//debug("Unexpected sibBase");
519 					return true;
520 			}
521 		} else {
522 			MCOperand_CreateReg0(mcInst, 0);
523 		}
524 
525 		// Check whether we are handling VSIB addressing mode for GATHER.
526 		// If sibIndex was set to SIB_INDEX_NONE, index offset is 4 and
527 		// we should use SIB_INDEX_XMM4|YMM4 for VSIB.
528 		// I don't see a way to get the correct IndexReg in readSIB:
529 		//   We can tell whether it is VSIB or SIB after instruction ID is decoded,
530 		//   but instruction ID may not be decoded yet when calling readSIB.
531 #ifndef CAPSTONE_X86_REDUCE
532 		Opcode = MCInst_getOpcode(mcInst);
533 #endif
534 		IndexIs128 = (
535 #ifndef CAPSTONE_X86_REDUCE
536 				Opcode == X86_VGATHERDPDrm ||
537 				Opcode == X86_VGATHERDPDYrm ||
538 				Opcode == X86_VGATHERQPDrm ||
539 				Opcode == X86_VGATHERDPSrm ||
540 				Opcode == X86_VGATHERQPSrm ||
541 				Opcode == X86_VPGATHERDQrm ||
542 				Opcode == X86_VPGATHERDQYrm ||
543 				Opcode == X86_VPGATHERQQrm ||
544 				Opcode == X86_VPGATHERDDrm ||
545 				Opcode == X86_VPGATHERQDrm ||
546 #endif
547 				false
548 				);
549 		IndexIs256 = (
550 #ifndef CAPSTONE_X86_REDUCE
551 				Opcode == X86_VGATHERQPDYrm ||
552 				Opcode == X86_VGATHERDPSYrm ||
553 				Opcode == X86_VGATHERQPSYrm ||
554 				Opcode == X86_VGATHERDPDZrm ||
555 				Opcode == X86_VPGATHERDQZrm ||
556 				Opcode == X86_VPGATHERQQYrm ||
557 				Opcode == X86_VPGATHERDDYrm ||
558 				Opcode == X86_VPGATHERQDYrm ||
559 #endif
560 				false
561 				);
562 		IndexIs512 = (
563 #ifndef CAPSTONE_X86_REDUCE
564 				Opcode == X86_VGATHERQPDZrm ||
565 				Opcode == X86_VGATHERDPSZrm ||
566 				Opcode == X86_VGATHERQPSZrm ||
567 				Opcode == X86_VPGATHERQQZrm ||
568 				Opcode == X86_VPGATHERDDZrm ||
569 				Opcode == X86_VPGATHERQDZrm ||
570 #endif
571 				false
572 				);
573 
574 		if (IndexIs128 || IndexIs256 || IndexIs512) {
575 			unsigned IndexOffset = insn->sibIndex -
576 				(insn->addressSize == 8 ? SIB_INDEX_RAX:SIB_INDEX_EAX);
577 			SIBIndex IndexBase = IndexIs512 ? SIB_INDEX_ZMM0 :
578 				IndexIs256 ? SIB_INDEX_YMM0 : SIB_INDEX_XMM0;
579 
580 			insn->sibIndex = (SIBIndex)(IndexBase + (insn->sibIndex == SIB_INDEX_NONE ? 4 : IndexOffset));
581 		}
582 
583 		if (insn->sibIndex != SIB_INDEX_NONE) {
584 			switch (insn->sibIndex) {
585 				default:
586 					//debug("Unexpected sibIndex");
587 					return true;
588 #define ENTRY(x)                                          \
589 				case SIB_INDEX_##x:                                 \
590 					indexReg = X86_##x; break;
591 					EA_BASES_32BIT
592 						EA_BASES_64BIT
593 						REGS_XMM
594 						REGS_YMM
595 						REGS_ZMM
596 #undef ENTRY
597 			}
598 		} else {
599 			indexReg = 0;
600 		}
601 
602 		scaleAmount = insn->sibScale;
603 	} else {
604 		switch (insn->eaBase) {
605 			case EA_BASE_NONE:
606 				if (insn->eaDisplacement == EA_DISP_NONE) {
607 					//debug("EA_BASE_NONE and EA_DISP_NONE for ModR/M base");
608 					return true;
609 				}
610 				if (insn->mode == MODE_64BIT) {
611 					if (insn->prefix3 == 0x67)	// address-size prefix overrides RIP relative addressing
612 						MCOperand_CreateReg0(mcInst, X86_EIP);
613 					else
614 						MCOperand_CreateReg0(mcInst, X86_RIP); // Section 2.2.1.6
615 				} else {
616 					MCOperand_CreateReg0(mcInst, 0);
617 				}
618 
619 				indexReg = 0;
620 				break;
621 			case EA_BASE_BX_SI:
622 				MCOperand_CreateReg0(mcInst, X86_BX);
623 				indexReg = X86_SI;
624 				break;
625 			case EA_BASE_BX_DI:
626 				MCOperand_CreateReg0(mcInst, X86_BX);
627 				indexReg = X86_DI;
628 				break;
629 			case EA_BASE_BP_SI:
630 				MCOperand_CreateReg0(mcInst, X86_BP);
631 				indexReg = X86_SI;
632 				break;
633 			case EA_BASE_BP_DI:
634 				MCOperand_CreateReg0(mcInst, X86_BP);
635 				indexReg = X86_DI;
636 				break;
637 			default:
638 				indexReg = 0;
639 				switch (insn->eaBase) {
640 					default:
641 						//debug("Unexpected eaBase");
642 						return true;
643 						// Here, we will use the fill-ins defined above.  However,
644 						//   BX_SI, BX_DI, BP_SI, and BP_DI are all handled above and
645 						//   sib and sib64 were handled in the top-level if, so they're only
646 						//   placeholders to keep the compiler happy.
647 #define ENTRY(x)                                        \
648 					case EA_BASE_##x:                                 \
649 						  MCOperand_CreateReg0(mcInst, X86_##x); break;
650 						ALL_EA_BASES
651 #undef ENTRY
652 #define ENTRY(x) case EA_REG_##x:
653 							ALL_REGS
654 #undef ENTRY
655 							//debug("A R/M memory operand may not be a register; "
656 							//      "the base field must be a base.");
657 							return true;
658 				}
659 		}
660 
661 		scaleAmount = 1;
662 	}
663 
664 	MCOperand_CreateImm0(mcInst, scaleAmount);
665 	MCOperand_CreateReg0(mcInst, indexReg);
666 	MCOperand_CreateImm0(mcInst, insn->displacement);
667 
668 	MCOperand_CreateReg0(mcInst, segmentRegnums[insn->segmentOverride]);
669 
670 	return false;
671 }
672 
673 /// translateRM - Translates an operand stored in the R/M (and possibly SIB)
674 ///   byte of an instruction to LLVM form, and appends it to an MCInst.
675 ///
676 /// @param mcInst       - The MCInst to append to.
677 /// @param operand      - The operand, as stored in the descriptor table.
678 /// @param insn         - The instruction to extract Mod, R/M, and SIB fields
679 ///                       from.
680 /// @return             - 0 on success; nonzero otherwise
translateRM(MCInst * mcInst,const OperandSpecifier * operand,InternalInstruction * insn)681 static bool translateRM(MCInst *mcInst, const OperandSpecifier *operand,
682 		InternalInstruction *insn)
683 {
684 	switch (operand->type) {
685 		case TYPE_R8:
686 		case TYPE_R16:
687 		case TYPE_R32:
688 		case TYPE_R64:
689 		case TYPE_Rv:
690 		case TYPE_MM64:
691 		case TYPE_XMM:
692 		case TYPE_XMM32:
693 		case TYPE_XMM64:
694 		case TYPE_XMM128:
695 		case TYPE_XMM256:
696 		case TYPE_XMM512:
697 		case TYPE_VK1:
698 		case TYPE_VK8:
699 		case TYPE_VK16:
700 		case TYPE_DEBUGREG:
701 		case TYPE_CONTROLREG:
702 			return translateRMRegister(mcInst, insn);
703 		case TYPE_M:
704 		case TYPE_M8:
705 		case TYPE_M16:
706 		case TYPE_M32:
707 		case TYPE_M64:
708 		case TYPE_M128:
709 		case TYPE_M256:
710 		case TYPE_M512:
711 		case TYPE_Mv:
712 		case TYPE_M32FP:
713 		case TYPE_M64FP:
714 		case TYPE_M80FP:
715 		case TYPE_M1616:
716 		case TYPE_M1632:
717 		case TYPE_M1664:
718 		case TYPE_LEA:
719 			return translateRMMemory(mcInst, insn);
720 		default:
721 			//debug("Unexpected type for a R/M operand");
722 			return true;
723 	}
724 }
725 
726 /// translateFPRegister - Translates a stack position on the FPU stack to its
727 ///   LLVM form, and appends it to an MCInst.
728 ///
729 /// @param mcInst       - The MCInst to append to.
730 /// @param stackPos     - The stack position to translate.
translateFPRegister(MCInst * mcInst,uint8_t stackPos)731 static void translateFPRegister(MCInst *mcInst, uint8_t stackPos)
732 {
733 	MCOperand_CreateReg0(mcInst, X86_ST0 + stackPos);
734 }
735 
736 /// translateMaskRegister - Translates a 3-bit mask register number to
737 ///   LLVM form, and appends it to an MCInst.
738 ///
739 /// @param mcInst       - The MCInst to append to.
740 /// @param maskRegNum   - Number of mask register from 0 to 7.
741 /// @return             - false on success; true otherwise.
translateMaskRegister(MCInst * mcInst,uint8_t maskRegNum)742 static bool translateMaskRegister(MCInst *mcInst, uint8_t maskRegNum)
743 {
744 	if (maskRegNum >= 8) {
745 		// debug("Invalid mask register number");
746 		return true;
747 	}
748 
749 	MCOperand_CreateReg0(mcInst, X86_K0 + maskRegNum);
750 
751 	return false;
752 }
753 
754 /// translateOperand - Translates an operand stored in an internal instruction
755 ///   to LLVM's format and appends it to an MCInst.
756 ///
757 /// @param mcInst       - The MCInst to append to.
758 /// @param operand      - The operand, as stored in the descriptor table.
759 /// @param insn         - The internal instruction.
760 /// @return             - false on success; true otherwise.
translateOperand(MCInst * mcInst,const OperandSpecifier * operand,InternalInstruction * insn)761 static bool translateOperand(MCInst *mcInst, const OperandSpecifier *operand, InternalInstruction *insn)
762 {
763 	switch (operand->encoding) {
764 		case ENCODING_REG:
765 			translateRegister(mcInst, insn->reg);
766 			return false;
767 		case ENCODING_WRITEMASK:
768 			return translateMaskRegister(mcInst, insn->writemask);
769 		CASE_ENCODING_RM:
770 			return translateRM(mcInst, operand, insn);
771 		case ENCODING_CB:
772 		case ENCODING_CW:
773 		case ENCODING_CD:
774 		case ENCODING_CP:
775 		case ENCODING_CO:
776 		case ENCODING_CT:
777 			//debug("Translation of code offsets isn't supported.");
778 			return true;
779 		case ENCODING_IB:
780 		case ENCODING_IW:
781 		case ENCODING_ID:
782 		case ENCODING_IO:
783 		case ENCODING_Iv:
784 		case ENCODING_Ia:
785 			translateImmediate(mcInst, insn->immediates[insn->numImmediatesTranslated++], operand, insn);
786 			return false;
787 		case ENCODING_SI:
788 			return translateSrcIndex(mcInst, insn);
789 		case ENCODING_DI:
790 			return translateDstIndex(mcInst, insn);
791 		case ENCODING_RB:
792 		case ENCODING_RW:
793 		case ENCODING_RD:
794 		case ENCODING_RO:
795 		case ENCODING_Rv:
796 			translateRegister(mcInst, insn->opcodeRegister);
797 			return false;
798 		case ENCODING_FP:
799 			translateFPRegister(mcInst, insn->modRM & 7);
800 			return false;
801 		case ENCODING_VVVV:
802 			translateRegister(mcInst, insn->vvvv);
803 			return false;
804 		case ENCODING_DUP:
805 			return translateOperand(mcInst, &insn->operands[operand->type - TYPE_DUP0], insn);
806 		default:
807 			//debug("Unhandled operand encoding during translation");
808 			return true;
809 	}
810 }
811 
translateInstruction(MCInst * mcInst,InternalInstruction * insn)812 static bool translateInstruction(MCInst *mcInst, InternalInstruction *insn)
813 {
814 	int index;
815 
816 	if (!insn->spec) {
817 		//debug("Instruction has no specification");
818 		return true;
819 	}
820 
821 	MCInst_setOpcode(mcInst, insn->instructionID);
822 
823 	// If when reading the prefix bytes we determined the overlapping 0xf2 or 0xf3
824 	// prefix bytes should be disassembled as xrelease and xacquire then set the
825 	// opcode to those instead of the rep and repne opcodes.
826 #ifndef CAPSTONE_X86_REDUCE
827 	if (insn->xAcquireRelease) {
828 		if (MCInst_getOpcode(mcInst) == X86_REP_PREFIX)
829 			MCInst_setOpcode(mcInst, X86_XRELEASE_PREFIX);
830 		else if (MCInst_getOpcode(mcInst) == X86_REPNE_PREFIX)
831 			MCInst_setOpcode(mcInst, X86_XACQUIRE_PREFIX);
832 	}
833 #endif
834 
835 	insn->numImmediatesTranslated = 0;
836 
837 	for (index = 0; index < X86_MAX_OPERANDS; ++index) {
838 		if (insn->operands[index].encoding != ENCODING_NONE) {
839 			if (translateOperand(mcInst, &insn->operands[index], insn)) {
840 				return true;
841 			}
842 		}
843 	}
844 
845 	return false;
846 }
847 
reader(const struct reader_info * info,uint8_t * byte,uint64_t address)848 static int reader(const struct reader_info *info, uint8_t *byte, uint64_t address)
849 {
850 	if (address - info->offset >= info->size)
851 		// out of buffer range
852 		return -1;
853 
854 	*byte = info->code[address - info->offset];
855 
856 	return 0;
857 }
858 
859 // copy x86 detail information from internal structure to public structure
update_pub_insn(cs_insn * pub,InternalInstruction * inter,uint8_t * prefixes)860 static void update_pub_insn(cs_insn *pub, InternalInstruction *inter, uint8_t *prefixes)
861 {
862 	prefixes[0] = inter->prefix0;
863 	prefixes[1] = inter->prefix1;
864 	prefixes[2] = inter->prefix2;
865 	prefixes[3] = inter->prefix3;
866 
867 	if (inter->vectorExtensionType != 0)
868 		memcpy(pub->detail->x86.opcode, inter->vectorExtensionPrefix, sizeof(pub->detail->x86.opcode));
869 	else {
870 		if (inter->twoByteEscape) {
871 			if (inter->threeByteEscape) {
872 				pub->detail->x86.opcode[0] = inter->twoByteEscape;
873 				pub->detail->x86.opcode[1] = inter->threeByteEscape;
874 				pub->detail->x86.opcode[2] = inter->opcode;
875 			} else {
876 				pub->detail->x86.opcode[0] = inter->twoByteEscape;
877 				pub->detail->x86.opcode[1] = inter->opcode;
878 			}
879 		} else {
880 				pub->detail->x86.opcode[0] = inter->opcode;
881 		}
882 	}
883 
884 	pub->detail->x86.rex = inter->rexPrefix;
885 
886 	pub->detail->x86.addr_size = inter->addressSize;
887 
888 	pub->detail->x86.modrm = inter->orgModRM;
889 	pub->detail->x86.encoding.modrm_offset = inter->modRMOffset;
890 
891 	pub->detail->x86.sib = inter->sib;
892 	pub->detail->x86.sib_index = x86_map_sib_index(inter->sibIndex);
893 	pub->detail->x86.sib_scale = inter->sibScale;
894 	pub->detail->x86.sib_base = x86_map_sib_base(inter->sibBase);
895 
896 	pub->detail->x86.disp = inter->displacement;
897 	if (inter->consumedDisplacement) {
898 		pub->detail->x86.encoding.disp_offset = inter->displacementOffset;
899 		pub->detail->x86.encoding.disp_size = inter->displacementSize;
900 	}
901 
902 	pub->detail->x86.encoding.imm_offset = inter->immediateOffset;
903 	if (pub->detail->x86.encoding.imm_size == 0 && inter->immediateOffset != 0)
904 		pub->detail->x86.encoding.imm_size = inter->immediateSize;
905 }
906 
X86_init(MCRegisterInfo * MRI)907 void X86_init(MCRegisterInfo *MRI)
908 {
909 	/*
910 	   InitMCRegisterInfo(X86RegDesc, 234,
911 	   RA, PC,
912 	   X86MCRegisterClasses, 79,
913 	   X86RegUnitRoots, 119, X86RegDiffLists, X86RegStrings,
914 	   X86SubRegIdxLists, 7,
915 	   X86SubRegIdxRanges, X86RegEncodingTable);
916 	*/
917 
918 	MCRegisterInfo_InitMCRegisterInfo(MRI, X86RegDesc, 234,
919 			0, 0,
920 			X86MCRegisterClasses, 79,
921 			0, 0, X86RegDiffLists, 0,
922 			X86SubRegIdxLists, 7,
923 			0);
924 }
925 
926 // Public interface for the disassembler
X86_getInstruction(csh ud,const uint8_t * code,size_t code_len,MCInst * instr,uint16_t * size,uint64_t address,void * _info)927 bool X86_getInstruction(csh ud, const uint8_t *code, size_t code_len,
928 		MCInst *instr, uint16_t *size, uint64_t address, void *_info)
929 {
930 	cs_struct *handle = (cs_struct *)(uintptr_t)ud;
931 	InternalInstruction insn = {0};
932 	struct reader_info info;
933 	int ret;
934 	bool result;
935 
936 	info.code = code;
937 	info.size = code_len;
938 	info.offset = address;
939 
940 	if (instr->flat_insn->detail) {
941 		// instr->flat_insn->detail initialization: 3 alternatives
942 
943 		// 1. The whole structure, this is how it's done in other arch disassemblers
944 		// Probably overkill since cs_detail is huge because of the 36 operands of ARM
945 
946 		//memset(instr->flat_insn->detail, 0, sizeof(cs_detail));
947 
948 		// 2. Only the part relevant to x86
949 		memset(instr->flat_insn->detail, 0, offsetof(cs_detail, x86) + sizeof(cs_x86));
950 
951 		// 3. The relevant part except for x86.operands
952 		// sizeof(cs_x86) is 0x1c0, sizeof(x86.operands) is 0x180
953 		// marginally faster, should be okay since x86.op_count is set to 0
954 
955 		//memset(instr->flat_insn->detail, 0, offsetof(cs_detail, x86)+offsetof(cs_x86, operands));
956 	}
957 
958 	if (handle->mode & CS_MODE_16)
959 		ret = decodeInstruction(&insn,
960 				reader, &info,
961 				address,
962 				MODE_16BIT);
963 	else if (handle->mode & CS_MODE_32)
964 		ret = decodeInstruction(&insn,
965 				reader, &info,
966 				address,
967 				MODE_32BIT);
968 	else
969 		ret = decodeInstruction(&insn,
970 				reader, &info,
971 				address,
972 				MODE_64BIT);
973 
974 	if (ret) {
975 		*size = (uint16_t)(insn.readerCursor - address);
976 		// handle some special cases here.
977 		// FIXME: fix this in the next major update.
978 		switch(*size) {
979 			default:
980 				break;
981 			case 2: {
982 						unsigned char b1 = 0, b2 = 0;
983 
984 						reader(&info, &b1, address);
985 						reader(&info, &b2, address + 1);
986 						if (b1 == 0x0f && b2 == 0xff) {
987 							instr->Opcode = X86_UD0;
988 							instr->OpcodePub = X86_INS_UD0;
989 							strncpy(instr->assembly, "ud0", 4);
990 							if (instr->flat_insn->detail) {
991 								instr->flat_insn->detail->x86.opcode[0] = b1;
992 								instr->flat_insn->detail->x86.opcode[1] = b2;
993 							}
994 							return true;
995 						}
996 				}
997 				return false;
998 			case 4: {
999 						if (handle->mode != CS_MODE_16) {
1000 							unsigned char b1 = 0, b2 = 0, b3 = 0, b4 = 0;
1001 
1002 							reader(&info, &b1, address);
1003 							reader(&info, &b2, address + 1);
1004 							reader(&info, &b3, address + 2);
1005 							reader(&info, &b4, address + 3);
1006 
1007 							if (b1 == 0xf3 && b2 == 0x0f && b3 == 0x1e && b4 == 0xfa) {
1008 								instr->Opcode = X86_ENDBR64;
1009 								instr->OpcodePub = X86_INS_ENDBR64;
1010 								strncpy(instr->assembly, "endbr64", 8);
1011 								if (instr->flat_insn->detail) {
1012 									instr->flat_insn->detail->x86.opcode[0] = b1;
1013 									instr->flat_insn->detail->x86.opcode[1] = b2;
1014 									instr->flat_insn->detail->x86.opcode[2] = b3;
1015 									instr->flat_insn->detail->x86.opcode[3] = b4;
1016 								}
1017 								return true;
1018 							} else if (b1 == 0xf3 && b2 == 0x0f && b3 == 0x1e && b4 == 0xfb) {
1019 								instr->Opcode = X86_ENDBR32;
1020 								instr->OpcodePub = X86_INS_ENDBR32;
1021 								strncpy(instr->assembly, "endbr32", 8);
1022 								if (instr->flat_insn->detail) {
1023 									instr->flat_insn->detail->x86.opcode[0] = b1;
1024 									instr->flat_insn->detail->x86.opcode[1] = b2;
1025 									instr->flat_insn->detail->x86.opcode[2] = b3;
1026 									instr->flat_insn->detail->x86.opcode[3] = b4;
1027 								}
1028 								return true;
1029 							}
1030 						}
1031 				}
1032 				return false;
1033 		}
1034 
1035 		return false;
1036 	} else {
1037 		*size = (uint16_t)insn.length;
1038 
1039 		result = (!translateInstruction(instr, &insn)) ?  true : false;
1040 		if (result) {
1041 			// quick fix for #904. TODO: fix this properly in the next update
1042 			if (handle->mode & CS_MODE_64) {
1043 				if (instr->Opcode == X86_LES16rm || instr->Opcode == X86_LES32rm)
1044 					// LES is invalid in x64
1045 					return false;
1046 				if (instr->Opcode == X86_LDS16rm || instr->Opcode == X86_LDS32rm)
1047 					// LDS is invalid in x64
1048 					return false;
1049 			}
1050 
1051 			instr->imm_size = insn.immSize;
1052 			if (handle->detail) {
1053 				update_pub_insn(instr->flat_insn, &insn, instr->x86_prefix);
1054 			} else {
1055 				// still copy all prefixes
1056 				instr->x86_prefix[0] = insn.prefix0;
1057 				instr->x86_prefix[1] = insn.prefix1;
1058 				instr->x86_prefix[2] = insn.prefix2;
1059 				instr->x86_prefix[3] = insn.prefix3;
1060 			}
1061 		}
1062 
1063 		return result;
1064 	}
1065 }
1066 
1067 #endif
1068