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 #include "X86Disassembler.h"
18 #include "X86DisassemblerDecoder.h"
19 #include "llvm/MC/MCContext.h"
20 #include "llvm/MC/MCDisassembler.h"
21 #include "llvm/MC/MCExpr.h"
22 #include "llvm/MC/MCInst.h"
23 #include "llvm/MC/MCInstrInfo.h"
24 #include "llvm/MC/MCSubtargetInfo.h"
25 #include "llvm/Support/Debug.h"
26 #include "llvm/Support/TargetRegistry.h"
27 #include "llvm/Support/raw_ostream.h"
28 
29 using namespace llvm;
30 using namespace llvm::X86Disassembler;
31 
32 #define DEBUG_TYPE "x86-disassembler"
33 
34 #define GET_REGINFO_ENUM
35 #include "X86GenRegisterInfo.inc"
36 #define GET_INSTRINFO_ENUM
37 #include "X86GenInstrInfo.inc"
38 #define GET_SUBTARGETINFO_ENUM
39 #include "X86GenSubtargetInfo.inc"
40 
Debug(const char * file,unsigned line,const char * s)41 void llvm::X86Disassembler::Debug(const char *file, unsigned line,
42                                   const char *s) {
43   dbgs() << file << ":" << line << ": " << s;
44 }
45 
GetInstrName(unsigned Opcode,const void * mii)46 const char *llvm::X86Disassembler::GetInstrName(unsigned Opcode,
47                                                 const void *mii) {
48   const MCInstrInfo *MII = static_cast<const MCInstrInfo *>(mii);
49   return MII->getName(Opcode);
50 }
51 
52 #define debug(s) DEBUG(Debug(__FILE__, __LINE__, s));
53 
54 namespace llvm {
55 
56 // Fill-ins to make the compiler happy.  These constants are never actually
57 //   assigned; they are just filler to make an automatically-generated switch
58 //   statement work.
59 namespace X86 {
60   enum {
61     BX_SI = 500,
62     BX_DI = 501,
63     BP_SI = 502,
64     BP_DI = 503,
65     sib   = 504,
66     sib64 = 505
67   };
68 }
69 
70 extern Target TheX86_32Target, TheX86_64Target;
71 
72 }
73 
74 static bool translateInstruction(MCInst &target,
75                                 InternalInstruction &source,
76                                 const MCDisassembler *Dis);
77 
X86GenericDisassembler(const MCSubtargetInfo & STI,MCContext & Ctx,std::unique_ptr<const MCInstrInfo> MII)78 X86GenericDisassembler::X86GenericDisassembler(
79                                          const MCSubtargetInfo &STI,
80                                          MCContext &Ctx,
81                                          std::unique_ptr<const MCInstrInfo> MII)
82   : MCDisassembler(STI, Ctx), MII(std::move(MII)) {
83   switch (STI.getFeatureBits() &
84           (X86::Mode16Bit | X86::Mode32Bit | X86::Mode64Bit)) {
85   case X86::Mode16Bit:
86     fMode = MODE_16BIT;
87     break;
88   case X86::Mode32Bit:
89     fMode = MODE_32BIT;
90     break;
91   case X86::Mode64Bit:
92     fMode = MODE_64BIT;
93     break;
94   default:
95     llvm_unreachable("Invalid CPU mode");
96   }
97 }
98 
99 struct Region {
100   ArrayRef<uint8_t> Bytes;
101   uint64_t Base;
RegionRegion102   Region(ArrayRef<uint8_t> Bytes, uint64_t Base) : Bytes(Bytes), Base(Base) {}
103 };
104 
105 /// A callback function that wraps the readByte method from Region.
106 ///
107 /// @param Arg      - The generic callback parameter.  In this case, this should
108 ///                   be a pointer to a Region.
109 /// @param Byte     - A pointer to the byte to be read.
110 /// @param Address  - The address to be read.
regionReader(const void * Arg,uint8_t * Byte,uint64_t Address)111 static int regionReader(const void *Arg, uint8_t *Byte, uint64_t Address) {
112   auto *R = static_cast<const Region *>(Arg);
113   ArrayRef<uint8_t> Bytes = R->Bytes;
114   unsigned Index = Address - R->Base;
115   if (Bytes.size() <= Index)
116     return -1;
117   *Byte = Bytes[Index];
118   return 0;
119 }
120 
121 /// logger - a callback function that wraps the operator<< method from
122 ///   raw_ostream.
123 ///
124 /// @param arg      - The generic callback parameter.  This should be a pointe
125 ///                   to a raw_ostream.
126 /// @param log      - A string to be logged.  logger() adds a newline.
logger(void * arg,const char * log)127 static void logger(void* arg, const char* log) {
128   if (!arg)
129     return;
130 
131   raw_ostream &vStream = *(static_cast<raw_ostream*>(arg));
132   vStream << log << "\n";
133 }
134 
135 //
136 // Public interface for the disassembler
137 //
138 
getInstruction(MCInst & Instr,uint64_t & Size,ArrayRef<uint8_t> Bytes,uint64_t Address,raw_ostream & VStream,raw_ostream & CStream) const139 MCDisassembler::DecodeStatus X86GenericDisassembler::getInstruction(
140     MCInst &Instr, uint64_t &Size, ArrayRef<uint8_t> Bytes, uint64_t Address,
141     raw_ostream &VStream, raw_ostream &CStream) const {
142   CommentStream = &CStream;
143 
144   InternalInstruction InternalInstr;
145 
146   dlog_t LoggerFn = logger;
147   if (&VStream == &nulls())
148     LoggerFn = nullptr; // Disable logging completely if it's going to nulls().
149 
150   Region R(Bytes, Address);
151 
152   int Ret = decodeInstruction(&InternalInstr, regionReader, (const void *)&R,
153                               LoggerFn, (void *)&VStream,
154                               (const void *)MII.get(), Address, fMode);
155 
156   if (Ret) {
157     Size = InternalInstr.readerCursor - Address;
158     return Fail;
159   } else {
160     Size = InternalInstr.length;
161     return (!translateInstruction(Instr, InternalInstr, this)) ? Success : Fail;
162   }
163 }
164 
165 //
166 // Private code that translates from struct InternalInstructions to MCInsts.
167 //
168 
169 /// translateRegister - Translates an internal register to the appropriate LLVM
170 ///   register, and appends it as an operand to an MCInst.
171 ///
172 /// @param mcInst     - The MCInst to append to.
173 /// @param reg        - The Reg to append.
translateRegister(MCInst & mcInst,Reg reg)174 static void translateRegister(MCInst &mcInst, Reg reg) {
175 #define ENTRY(x) X86::x,
176   uint8_t llvmRegnums[] = {
177     ALL_REGS
178     0
179   };
180 #undef ENTRY
181 
182   uint8_t llvmRegnum = llvmRegnums[reg];
183   mcInst.addOperand(MCOperand::CreateReg(llvmRegnum));
184 }
185 
186 /// tryAddingSymbolicOperand - trys to add a symbolic operand in place of the
187 /// immediate Value in the MCInst.
188 ///
189 /// @param Value      - The immediate Value, has had any PC adjustment made by
190 ///                     the caller.
191 /// @param isBranch   - If the instruction is a branch instruction
192 /// @param Address    - The starting address of the instruction
193 /// @param Offset     - The byte offset to this immediate in the instruction
194 /// @param Width      - The byte width of this immediate in the instruction
195 ///
196 /// If the getOpInfo() function was set when setupForSymbolicDisassembly() was
197 /// called then that function is called to get any symbolic information for the
198 /// immediate in the instruction using the Address, Offset and Width.  If that
199 /// returns non-zero then the symbolic information it returns is used to create
200 /// an MCExpr and that is added as an operand to the MCInst.  If getOpInfo()
201 /// returns zero and isBranch is true then a symbol look up for immediate Value
202 /// is done and if a symbol is found an MCExpr is created with that, else
203 /// an MCExpr with the immediate Value is created.  This function returns true
204 /// if it adds an operand to the MCInst and false otherwise.
tryAddingSymbolicOperand(int64_t Value,bool isBranch,uint64_t Address,uint64_t Offset,uint64_t Width,MCInst & MI,const MCDisassembler * Dis)205 static bool tryAddingSymbolicOperand(int64_t Value, bool isBranch,
206                                      uint64_t Address, uint64_t Offset,
207                                      uint64_t Width, MCInst &MI,
208                                      const MCDisassembler *Dis) {
209   return Dis->tryAddingSymbolicOperand(MI, Value, Address, isBranch,
210                                        Offset, Width);
211 }
212 
213 /// tryAddingPcLoadReferenceComment - trys to add a comment as to what is being
214 /// referenced by a load instruction with the base register that is the rip.
215 /// These can often be addresses in a literal pool.  The Address of the
216 /// instruction and its immediate Value are used to determine the address
217 /// being referenced in the literal pool entry.  The SymbolLookUp call back will
218 /// return a pointer to a literal 'C' string if the referenced address is an
219 /// address into a section with 'C' string literals.
tryAddingPcLoadReferenceComment(uint64_t Address,uint64_t Value,const void * Decoder)220 static void tryAddingPcLoadReferenceComment(uint64_t Address, uint64_t Value,
221                                             const void *Decoder) {
222   const MCDisassembler *Dis = static_cast<const MCDisassembler*>(Decoder);
223   Dis->tryAddingPcLoadReferenceComment(Value, Address);
224 }
225 
226 static const uint8_t segmentRegnums[SEG_OVERRIDE_max] = {
227   0,        // SEG_OVERRIDE_NONE
228   X86::CS,
229   X86::SS,
230   X86::DS,
231   X86::ES,
232   X86::FS,
233   X86::GS
234 };
235 
236 /// translateSrcIndex   - Appends a source index operand to an MCInst.
237 ///
238 /// @param mcInst       - The MCInst to append to.
239 /// @param insn         - The internal instruction.
translateSrcIndex(MCInst & mcInst,InternalInstruction & insn)240 static bool translateSrcIndex(MCInst &mcInst, InternalInstruction &insn) {
241   unsigned baseRegNo;
242 
243   if (insn.mode == MODE_64BIT)
244     baseRegNo = insn.prefixPresent[0x67] ? X86::ESI : X86::RSI;
245   else if (insn.mode == MODE_32BIT)
246     baseRegNo = insn.prefixPresent[0x67] ? X86::SI : X86::ESI;
247   else {
248     assert(insn.mode == MODE_16BIT);
249     baseRegNo = insn.prefixPresent[0x67] ? X86::ESI : X86::SI;
250   }
251   MCOperand baseReg = MCOperand::CreateReg(baseRegNo);
252   mcInst.addOperand(baseReg);
253 
254   MCOperand segmentReg;
255   segmentReg = MCOperand::CreateReg(segmentRegnums[insn.segmentOverride]);
256   mcInst.addOperand(segmentReg);
257   return false;
258 }
259 
260 /// translateDstIndex   - Appends a destination index operand to an MCInst.
261 ///
262 /// @param mcInst       - The MCInst to append to.
263 /// @param insn         - The internal instruction.
264 
translateDstIndex(MCInst & mcInst,InternalInstruction & insn)265 static bool translateDstIndex(MCInst &mcInst, InternalInstruction &insn) {
266   unsigned baseRegNo;
267 
268   if (insn.mode == MODE_64BIT)
269     baseRegNo = insn.prefixPresent[0x67] ? X86::EDI : X86::RDI;
270   else if (insn.mode == MODE_32BIT)
271     baseRegNo = insn.prefixPresent[0x67] ? X86::DI : X86::EDI;
272   else {
273     assert(insn.mode == MODE_16BIT);
274     baseRegNo = insn.prefixPresent[0x67] ? X86::EDI : X86::DI;
275   }
276   MCOperand baseReg = MCOperand::CreateReg(baseRegNo);
277   mcInst.addOperand(baseReg);
278   return false;
279 }
280 
281 /// translateImmediate  - Appends an immediate operand to an MCInst.
282 ///
283 /// @param mcInst       - The MCInst to append to.
284 /// @param immediate    - The immediate value to append.
285 /// @param operand      - The operand, as stored in the descriptor table.
286 /// @param insn         - The internal instruction.
translateImmediate(MCInst & mcInst,uint64_t immediate,const OperandSpecifier & operand,InternalInstruction & insn,const MCDisassembler * Dis)287 static void translateImmediate(MCInst &mcInst, uint64_t immediate,
288                                const OperandSpecifier &operand,
289                                InternalInstruction &insn,
290                                const MCDisassembler *Dis) {
291   // Sign-extend the immediate if necessary.
292 
293   OperandType type = (OperandType)operand.type;
294 
295   bool isBranch = false;
296   uint64_t pcrel = 0;
297   if (type == TYPE_RELv) {
298     isBranch = true;
299     pcrel = insn.startLocation +
300             insn.immediateOffset + insn.immediateSize;
301     switch (insn.displacementSize) {
302     default:
303       break;
304     case 1:
305       if(immediate & 0x80)
306         immediate |= ~(0xffull);
307       break;
308     case 2:
309       if(immediate & 0x8000)
310         immediate |= ~(0xffffull);
311       break;
312     case 4:
313       if(immediate & 0x80000000)
314         immediate |= ~(0xffffffffull);
315       break;
316     case 8:
317       break;
318     }
319   }
320   // By default sign-extend all X86 immediates based on their encoding.
321   else if (type == TYPE_IMM8 || type == TYPE_IMM16 || type == TYPE_IMM32 ||
322            type == TYPE_IMM64 || type == TYPE_IMMv) {
323     switch (operand.encoding) {
324     default:
325       break;
326     case ENCODING_IB:
327       if(immediate & 0x80)
328         immediate |= ~(0xffull);
329       break;
330     case ENCODING_IW:
331       if(immediate & 0x8000)
332         immediate |= ~(0xffffull);
333       break;
334     case ENCODING_ID:
335       if(immediate & 0x80000000)
336         immediate |= ~(0xffffffffull);
337       break;
338     case ENCODING_IO:
339       break;
340     }
341   } else if (type == TYPE_IMM3) {
342     // Check for immediates that printSSECC can't handle.
343     if (immediate >= 8) {
344       unsigned NewOpc;
345       switch (mcInst.getOpcode()) {
346       default: llvm_unreachable("unexpected opcode");
347       case X86::CMPPDrmi:  NewOpc = X86::CMPPDrmi_alt;  break;
348       case X86::CMPPDrri:  NewOpc = X86::CMPPDrri_alt;  break;
349       case X86::CMPPSrmi:  NewOpc = X86::CMPPSrmi_alt;  break;
350       case X86::CMPPSrri:  NewOpc = X86::CMPPSrri_alt;  break;
351       case X86::CMPSDrm:   NewOpc = X86::CMPSDrm_alt;   break;
352       case X86::CMPSDrr:   NewOpc = X86::CMPSDrr_alt;   break;
353       case X86::CMPSSrm:   NewOpc = X86::CMPSSrm_alt;   break;
354       case X86::CMPSSrr:   NewOpc = X86::CMPSSrr_alt;   break;
355       case X86::VPCOMBri:  NewOpc = X86::VPCOMBri_alt;  break;
356       case X86::VPCOMBmi:  NewOpc = X86::VPCOMBmi_alt;  break;
357       case X86::VPCOMWri:  NewOpc = X86::VPCOMWri_alt;  break;
358       case X86::VPCOMWmi:  NewOpc = X86::VPCOMWmi_alt;  break;
359       case X86::VPCOMDri:  NewOpc = X86::VPCOMDri_alt;  break;
360       case X86::VPCOMDmi:  NewOpc = X86::VPCOMDmi_alt;  break;
361       case X86::VPCOMQri:  NewOpc = X86::VPCOMQri_alt;  break;
362       case X86::VPCOMQmi:  NewOpc = X86::VPCOMQmi_alt;  break;
363       case X86::VPCOMUBri: NewOpc = X86::VPCOMUBri_alt; break;
364       case X86::VPCOMUBmi: NewOpc = X86::VPCOMUBmi_alt; break;
365       case X86::VPCOMUWri: NewOpc = X86::VPCOMUWri_alt; break;
366       case X86::VPCOMUWmi: NewOpc = X86::VPCOMUWmi_alt; break;
367       case X86::VPCOMUDri: NewOpc = X86::VPCOMUDri_alt; break;
368       case X86::VPCOMUDmi: NewOpc = X86::VPCOMUDmi_alt; break;
369       case X86::VPCOMUQri: NewOpc = X86::VPCOMUQri_alt; break;
370       case X86::VPCOMUQmi: NewOpc = X86::VPCOMUQmi_alt; break;
371       }
372       // Switch opcode to the one that doesn't get special printing.
373       mcInst.setOpcode(NewOpc);
374     }
375   } else if (type == TYPE_IMM5) {
376     // Check for immediates that printAVXCC can't handle.
377     if (immediate >= 32) {
378       unsigned NewOpc;
379       switch (mcInst.getOpcode()) {
380       default: llvm_unreachable("unexpected opcode");
381       case X86::VCMPPDrmi:   NewOpc = X86::VCMPPDrmi_alt;   break;
382       case X86::VCMPPDrri:   NewOpc = X86::VCMPPDrri_alt;   break;
383       case X86::VCMPPSrmi:   NewOpc = X86::VCMPPSrmi_alt;   break;
384       case X86::VCMPPSrri:   NewOpc = X86::VCMPPSrri_alt;   break;
385       case X86::VCMPSDrm:    NewOpc = X86::VCMPSDrm_alt;    break;
386       case X86::VCMPSDrr:    NewOpc = X86::VCMPSDrr_alt;    break;
387       case X86::VCMPSSrm:    NewOpc = X86::VCMPSSrm_alt;    break;
388       case X86::VCMPSSrr:    NewOpc = X86::VCMPSSrr_alt;    break;
389       case X86::VCMPPDYrmi:  NewOpc = X86::VCMPPDYrmi_alt;  break;
390       case X86::VCMPPDYrri:  NewOpc = X86::VCMPPDYrri_alt;  break;
391       case X86::VCMPPSYrmi:  NewOpc = X86::VCMPPSYrmi_alt;  break;
392       case X86::VCMPPSYrri:  NewOpc = X86::VCMPPSYrri_alt;  break;
393       case X86::VCMPPDZrmi:  NewOpc = X86::VCMPPDZrmi_alt;  break;
394       case X86::VCMPPDZrri:  NewOpc = X86::VCMPPDZrri_alt;  break;
395       case X86::VCMPPDZrrib: NewOpc = X86::VCMPPDZrrib_alt; break;
396       case X86::VCMPPSZrmi:  NewOpc = X86::VCMPPSZrmi_alt;  break;
397       case X86::VCMPPSZrri:  NewOpc = X86::VCMPPSZrri_alt;  break;
398       case X86::VCMPPSZrrib: NewOpc = X86::VCMPPSZrrib_alt; break;
399       case X86::VCMPSDZrm:   NewOpc = X86::VCMPSDZrmi_alt;  break;
400       case X86::VCMPSDZrr:   NewOpc = X86::VCMPSDZrri_alt;  break;
401       case X86::VCMPSSZrm:   NewOpc = X86::VCMPSSZrmi_alt;  break;
402       case X86::VCMPSSZrr:   NewOpc = X86::VCMPSSZrri_alt;  break;
403       }
404       // Switch opcode to the one that doesn't get special printing.
405       mcInst.setOpcode(NewOpc);
406     }
407   } else if (type == TYPE_AVX512ICC) {
408     if (immediate >= 8 || ((immediate & 0x3) == 3)) {
409       unsigned NewOpc;
410       switch (mcInst.getOpcode()) {
411       default: llvm_unreachable("unexpected opcode");
412       case X86::VPCMPBZ128rmi:    NewOpc = X86::VPCMPBZ128rmi_alt;    break;
413       case X86::VPCMPBZ128rmik:   NewOpc = X86::VPCMPBZ128rmik_alt;   break;
414       case X86::VPCMPBZ128rri:    NewOpc = X86::VPCMPBZ128rri_alt;    break;
415       case X86::VPCMPBZ128rrik:   NewOpc = X86::VPCMPBZ128rrik_alt;   break;
416       case X86::VPCMPBZ256rmi:    NewOpc = X86::VPCMPBZ256rmi_alt;    break;
417       case X86::VPCMPBZ256rmik:   NewOpc = X86::VPCMPBZ256rmik_alt;   break;
418       case X86::VPCMPBZ256rri:    NewOpc = X86::VPCMPBZ256rri_alt;    break;
419       case X86::VPCMPBZ256rrik:   NewOpc = X86::VPCMPBZ256rrik_alt;   break;
420       case X86::VPCMPBZrmi:       NewOpc = X86::VPCMPBZrmi_alt;       break;
421       case X86::VPCMPBZrmik:      NewOpc = X86::VPCMPBZrmik_alt;      break;
422       case X86::VPCMPBZrri:       NewOpc = X86::VPCMPBZrri_alt;       break;
423       case X86::VPCMPBZrrik:      NewOpc = X86::VPCMPBZrrik_alt;      break;
424       case X86::VPCMPDZ128rmi:    NewOpc = X86::VPCMPDZ128rmi_alt;    break;
425       case X86::VPCMPDZ128rmib:   NewOpc = X86::VPCMPDZ128rmib_alt;   break;
426       case X86::VPCMPDZ128rmibk:  NewOpc = X86::VPCMPDZ128rmibk_alt;  break;
427       case X86::VPCMPDZ128rmik:   NewOpc = X86::VPCMPDZ128rmik_alt;   break;
428       case X86::VPCMPDZ128rri:    NewOpc = X86::VPCMPDZ128rri_alt;    break;
429       case X86::VPCMPDZ128rrik:   NewOpc = X86::VPCMPDZ128rrik_alt;   break;
430       case X86::VPCMPDZ256rmi:    NewOpc = X86::VPCMPDZ256rmi_alt;    break;
431       case X86::VPCMPDZ256rmib:   NewOpc = X86::VPCMPDZ256rmib_alt;   break;
432       case X86::VPCMPDZ256rmibk:  NewOpc = X86::VPCMPDZ256rmibk_alt;  break;
433       case X86::VPCMPDZ256rmik:   NewOpc = X86::VPCMPDZ256rmik_alt;   break;
434       case X86::VPCMPDZ256rri:    NewOpc = X86::VPCMPDZ256rri_alt;    break;
435       case X86::VPCMPDZ256rrik:   NewOpc = X86::VPCMPDZ256rrik_alt;   break;
436       case X86::VPCMPDZrmi:       NewOpc = X86::VPCMPDZrmi_alt;       break;
437       case X86::VPCMPDZrmib:      NewOpc = X86::VPCMPDZrmib_alt;      break;
438       case X86::VPCMPDZrmibk:     NewOpc = X86::VPCMPDZrmibk_alt;     break;
439       case X86::VPCMPDZrmik:      NewOpc = X86::VPCMPDZrmik_alt;      break;
440       case X86::VPCMPDZrri:       NewOpc = X86::VPCMPDZrri_alt;       break;
441       case X86::VPCMPDZrrik:      NewOpc = X86::VPCMPDZrrik_alt;      break;
442       case X86::VPCMPQZ128rmi:    NewOpc = X86::VPCMPQZ128rmi_alt;    break;
443       case X86::VPCMPQZ128rmib:   NewOpc = X86::VPCMPQZ128rmib_alt;   break;
444       case X86::VPCMPQZ128rmibk:  NewOpc = X86::VPCMPQZ128rmibk_alt;  break;
445       case X86::VPCMPQZ128rmik:   NewOpc = X86::VPCMPQZ128rmik_alt;   break;
446       case X86::VPCMPQZ128rri:    NewOpc = X86::VPCMPQZ128rri_alt;    break;
447       case X86::VPCMPQZ128rrik:   NewOpc = X86::VPCMPQZ128rrik_alt;   break;
448       case X86::VPCMPQZ256rmi:    NewOpc = X86::VPCMPQZ256rmi_alt;    break;
449       case X86::VPCMPQZ256rmib:   NewOpc = X86::VPCMPQZ256rmib_alt;   break;
450       case X86::VPCMPQZ256rmibk:  NewOpc = X86::VPCMPQZ256rmibk_alt;  break;
451       case X86::VPCMPQZ256rmik:   NewOpc = X86::VPCMPQZ256rmik_alt;   break;
452       case X86::VPCMPQZ256rri:    NewOpc = X86::VPCMPQZ256rri_alt;    break;
453       case X86::VPCMPQZ256rrik:   NewOpc = X86::VPCMPQZ256rrik_alt;   break;
454       case X86::VPCMPQZrmi:       NewOpc = X86::VPCMPQZrmi_alt;       break;
455       case X86::VPCMPQZrmib:      NewOpc = X86::VPCMPQZrmib_alt;      break;
456       case X86::VPCMPQZrmibk:     NewOpc = X86::VPCMPQZrmibk_alt;     break;
457       case X86::VPCMPQZrmik:      NewOpc = X86::VPCMPQZrmik_alt;      break;
458       case X86::VPCMPQZrri:       NewOpc = X86::VPCMPQZrri_alt;       break;
459       case X86::VPCMPQZrrik:      NewOpc = X86::VPCMPQZrrik_alt;      break;
460       case X86::VPCMPUBZ128rmi:   NewOpc = X86::VPCMPUBZ128rmi_alt;   break;
461       case X86::VPCMPUBZ128rmik:  NewOpc = X86::VPCMPUBZ128rmik_alt;  break;
462       case X86::VPCMPUBZ128rri:   NewOpc = X86::VPCMPUBZ128rri_alt;   break;
463       case X86::VPCMPUBZ128rrik:  NewOpc = X86::VPCMPUBZ128rrik_alt;  break;
464       case X86::VPCMPUBZ256rmi:   NewOpc = X86::VPCMPUBZ256rmi_alt;   break;
465       case X86::VPCMPUBZ256rmik:  NewOpc = X86::VPCMPUBZ256rmik_alt;  break;
466       case X86::VPCMPUBZ256rri:   NewOpc = X86::VPCMPUBZ256rri_alt;   break;
467       case X86::VPCMPUBZ256rrik:  NewOpc = X86::VPCMPUBZ256rrik_alt;  break;
468       case X86::VPCMPUBZrmi:      NewOpc = X86::VPCMPUBZrmi_alt;      break;
469       case X86::VPCMPUBZrmik:     NewOpc = X86::VPCMPUBZrmik_alt;     break;
470       case X86::VPCMPUBZrri:      NewOpc = X86::VPCMPUBZrri_alt;      break;
471       case X86::VPCMPUBZrrik:     NewOpc = X86::VPCMPUBZrrik_alt;     break;
472       case X86::VPCMPUDZ128rmi:   NewOpc = X86::VPCMPUDZ128rmi_alt;   break;
473       case X86::VPCMPUDZ128rmib:  NewOpc = X86::VPCMPUDZ128rmib_alt;  break;
474       case X86::VPCMPUDZ128rmibk: NewOpc = X86::VPCMPUDZ128rmibk_alt; break;
475       case X86::VPCMPUDZ128rmik:  NewOpc = X86::VPCMPUDZ128rmik_alt;  break;
476       case X86::VPCMPUDZ128rri:   NewOpc = X86::VPCMPUDZ128rri_alt;   break;
477       case X86::VPCMPUDZ128rrik:  NewOpc = X86::VPCMPUDZ128rrik_alt;  break;
478       case X86::VPCMPUDZ256rmi:   NewOpc = X86::VPCMPUDZ256rmi_alt;   break;
479       case X86::VPCMPUDZ256rmib:  NewOpc = X86::VPCMPUDZ256rmib_alt;  break;
480       case X86::VPCMPUDZ256rmibk: NewOpc = X86::VPCMPUDZ256rmibk_alt; break;
481       case X86::VPCMPUDZ256rmik:  NewOpc = X86::VPCMPUDZ256rmik_alt;  break;
482       case X86::VPCMPUDZ256rri:   NewOpc = X86::VPCMPUDZ256rri_alt;   break;
483       case X86::VPCMPUDZ256rrik:  NewOpc = X86::VPCMPUDZ256rrik_alt;  break;
484       case X86::VPCMPUDZrmi:      NewOpc = X86::VPCMPUDZrmi_alt;      break;
485       case X86::VPCMPUDZrmib:     NewOpc = X86::VPCMPUDZrmib_alt;     break;
486       case X86::VPCMPUDZrmibk:    NewOpc = X86::VPCMPUDZrmibk_alt;    break;
487       case X86::VPCMPUDZrmik:     NewOpc = X86::VPCMPUDZrmik_alt;     break;
488       case X86::VPCMPUDZrri:      NewOpc = X86::VPCMPUDZrri_alt;      break;
489       case X86::VPCMPUDZrrik:     NewOpc = X86::VPCMPUDZrrik_alt;     break;
490       case X86::VPCMPUQZ128rmi:   NewOpc = X86::VPCMPUQZ128rmi_alt;   break;
491       case X86::VPCMPUQZ128rmib:  NewOpc = X86::VPCMPUQZ128rmib_alt;  break;
492       case X86::VPCMPUQZ128rmibk: NewOpc = X86::VPCMPUQZ128rmibk_alt; break;
493       case X86::VPCMPUQZ128rmik:  NewOpc = X86::VPCMPUQZ128rmik_alt;  break;
494       case X86::VPCMPUQZ128rri:   NewOpc = X86::VPCMPUQZ128rri_alt;   break;
495       case X86::VPCMPUQZ128rrik:  NewOpc = X86::VPCMPUQZ128rrik_alt;  break;
496       case X86::VPCMPUQZ256rmi:   NewOpc = X86::VPCMPUQZ256rmi_alt;   break;
497       case X86::VPCMPUQZ256rmib:  NewOpc = X86::VPCMPUQZ256rmib_alt;  break;
498       case X86::VPCMPUQZ256rmibk: NewOpc = X86::VPCMPUQZ256rmibk_alt; break;
499       case X86::VPCMPUQZ256rmik:  NewOpc = X86::VPCMPUQZ256rmik_alt;  break;
500       case X86::VPCMPUQZ256rri:   NewOpc = X86::VPCMPUQZ256rri_alt;   break;
501       case X86::VPCMPUQZ256rrik:  NewOpc = X86::VPCMPUQZ256rrik_alt;  break;
502       case X86::VPCMPUQZrmi:      NewOpc = X86::VPCMPUQZrmi_alt;      break;
503       case X86::VPCMPUQZrmib:     NewOpc = X86::VPCMPUQZrmib_alt;     break;
504       case X86::VPCMPUQZrmibk:    NewOpc = X86::VPCMPUQZrmibk_alt;    break;
505       case X86::VPCMPUQZrmik:     NewOpc = X86::VPCMPUQZrmik_alt;     break;
506       case X86::VPCMPUQZrri:      NewOpc = X86::VPCMPUQZrri_alt;      break;
507       case X86::VPCMPUQZrrik:     NewOpc = X86::VPCMPUQZrrik_alt;     break;
508       case X86::VPCMPUWZ128rmi:   NewOpc = X86::VPCMPUWZ128rmi_alt;   break;
509       case X86::VPCMPUWZ128rmik:  NewOpc = X86::VPCMPUWZ128rmik_alt;  break;
510       case X86::VPCMPUWZ128rri:   NewOpc = X86::VPCMPUWZ128rri_alt;   break;
511       case X86::VPCMPUWZ128rrik:  NewOpc = X86::VPCMPUWZ128rrik_alt;  break;
512       case X86::VPCMPUWZ256rmi:   NewOpc = X86::VPCMPUWZ256rmi_alt;   break;
513       case X86::VPCMPUWZ256rmik:  NewOpc = X86::VPCMPUWZ256rmik_alt;  break;
514       case X86::VPCMPUWZ256rri:   NewOpc = X86::VPCMPUWZ256rri_alt;   break;
515       case X86::VPCMPUWZ256rrik:  NewOpc = X86::VPCMPUWZ256rrik_alt;  break;
516       case X86::VPCMPUWZrmi:      NewOpc = X86::VPCMPUWZrmi_alt;      break;
517       case X86::VPCMPUWZrmik:     NewOpc = X86::VPCMPUWZrmik_alt;     break;
518       case X86::VPCMPUWZrri:      NewOpc = X86::VPCMPUWZrri_alt;      break;
519       case X86::VPCMPUWZrrik:     NewOpc = X86::VPCMPUWZrrik_alt;     break;
520       case X86::VPCMPWZ128rmi:    NewOpc = X86::VPCMPWZ128rmi_alt;    break;
521       case X86::VPCMPWZ128rmik:   NewOpc = X86::VPCMPWZ128rmik_alt;   break;
522       case X86::VPCMPWZ128rri:    NewOpc = X86::VPCMPWZ128rri_alt;    break;
523       case X86::VPCMPWZ128rrik:   NewOpc = X86::VPCMPWZ128rrik_alt;   break;
524       case X86::VPCMPWZ256rmi:    NewOpc = X86::VPCMPWZ256rmi_alt;    break;
525       case X86::VPCMPWZ256rmik:   NewOpc = X86::VPCMPWZ256rmik_alt;   break;
526       case X86::VPCMPWZ256rri:    NewOpc = X86::VPCMPWZ256rri_alt;    break;
527       case X86::VPCMPWZ256rrik:   NewOpc = X86::VPCMPWZ256rrik_alt;   break;
528       case X86::VPCMPWZrmi:       NewOpc = X86::VPCMPWZrmi_alt;       break;
529       case X86::VPCMPWZrmik:      NewOpc = X86::VPCMPWZrmik_alt;      break;
530       case X86::VPCMPWZrri:       NewOpc = X86::VPCMPWZrri_alt;       break;
531       case X86::VPCMPWZrrik:      NewOpc = X86::VPCMPWZrrik_alt;      break;
532       }
533       // Switch opcode to the one that doesn't get special printing.
534       mcInst.setOpcode(NewOpc);
535     }
536   }
537 
538   switch (type) {
539   case TYPE_XMM32:
540   case TYPE_XMM64:
541   case TYPE_XMM128:
542     mcInst.addOperand(MCOperand::CreateReg(X86::XMM0 + (immediate >> 4)));
543     return;
544   case TYPE_XMM256:
545     mcInst.addOperand(MCOperand::CreateReg(X86::YMM0 + (immediate >> 4)));
546     return;
547   case TYPE_XMM512:
548     mcInst.addOperand(MCOperand::CreateReg(X86::ZMM0 + (immediate >> 4)));
549     return;
550   case TYPE_REL8:
551     isBranch = true;
552     pcrel = insn.startLocation + insn.immediateOffset + insn.immediateSize;
553     if(immediate & 0x80)
554       immediate |= ~(0xffull);
555     break;
556   case TYPE_REL32:
557   case TYPE_REL64:
558     isBranch = true;
559     pcrel = insn.startLocation + insn.immediateOffset + insn.immediateSize;
560     if(immediate & 0x80000000)
561       immediate |= ~(0xffffffffull);
562     break;
563   default:
564     // operand is 64 bits wide.  Do nothing.
565     break;
566   }
567 
568   if(!tryAddingSymbolicOperand(immediate + pcrel, isBranch, insn.startLocation,
569                                insn.immediateOffset, insn.immediateSize,
570                                mcInst, Dis))
571     mcInst.addOperand(MCOperand::CreateImm(immediate));
572 
573   if (type == TYPE_MOFFS8 || type == TYPE_MOFFS16 ||
574       type == TYPE_MOFFS32 || type == TYPE_MOFFS64) {
575     MCOperand segmentReg;
576     segmentReg = MCOperand::CreateReg(segmentRegnums[insn.segmentOverride]);
577     mcInst.addOperand(segmentReg);
578   }
579 }
580 
581 /// translateRMRegister - Translates a register stored in the R/M field of the
582 ///   ModR/M byte to its LLVM equivalent and appends it to an MCInst.
583 /// @param mcInst       - The MCInst to append to.
584 /// @param insn         - The internal instruction to extract the R/M field
585 ///                       from.
586 /// @return             - 0 on success; -1 otherwise
translateRMRegister(MCInst & mcInst,InternalInstruction & insn)587 static bool translateRMRegister(MCInst &mcInst,
588                                 InternalInstruction &insn) {
589   if (insn.eaBase == EA_BASE_sib || insn.eaBase == EA_BASE_sib64) {
590     debug("A R/M register operand may not have a SIB byte");
591     return true;
592   }
593 
594   switch (insn.eaBase) {
595   default:
596     debug("Unexpected EA base register");
597     return true;
598   case EA_BASE_NONE:
599     debug("EA_BASE_NONE for ModR/M base");
600     return true;
601 #define ENTRY(x) case EA_BASE_##x:
602   ALL_EA_BASES
603 #undef ENTRY
604     debug("A R/M register operand may not have a base; "
605           "the operand must be a register.");
606     return true;
607 #define ENTRY(x)                                                      \
608   case EA_REG_##x:                                                    \
609     mcInst.addOperand(MCOperand::CreateReg(X86::x)); break;
610   ALL_REGS
611 #undef ENTRY
612   }
613 
614   return false;
615 }
616 
617 /// translateRMMemory - Translates a memory operand stored in the Mod and R/M
618 ///   fields of an internal instruction (and possibly its SIB byte) to a memory
619 ///   operand in LLVM's format, and appends it to an MCInst.
620 ///
621 /// @param mcInst       - The MCInst to append to.
622 /// @param insn         - The instruction to extract Mod, R/M, and SIB fields
623 ///                       from.
624 /// @return             - 0 on success; nonzero otherwise
translateRMMemory(MCInst & mcInst,InternalInstruction & insn,const MCDisassembler * Dis)625 static bool translateRMMemory(MCInst &mcInst, InternalInstruction &insn,
626                               const MCDisassembler *Dis) {
627   // Addresses in an MCInst are represented as five operands:
628   //   1. basereg       (register)  The R/M base, or (if there is a SIB) the
629   //                                SIB base
630   //   2. scaleamount   (immediate) 1, or (if there is a SIB) the specified
631   //                                scale amount
632   //   3. indexreg      (register)  x86_registerNONE, or (if there is a SIB)
633   //                                the index (which is multiplied by the
634   //                                scale amount)
635   //   4. displacement  (immediate) 0, or the displacement if there is one
636   //   5. segmentreg    (register)  x86_registerNONE for now, but could be set
637   //                                if we have segment overrides
638 
639   MCOperand baseReg;
640   MCOperand scaleAmount;
641   MCOperand indexReg;
642   MCOperand displacement;
643   MCOperand segmentReg;
644   uint64_t pcrel = 0;
645 
646   if (insn.eaBase == EA_BASE_sib || insn.eaBase == EA_BASE_sib64) {
647     if (insn.sibBase != SIB_BASE_NONE) {
648       switch (insn.sibBase) {
649       default:
650         debug("Unexpected sibBase");
651         return true;
652 #define ENTRY(x)                                          \
653       case SIB_BASE_##x:                                  \
654         baseReg = MCOperand::CreateReg(X86::x); break;
655       ALL_SIB_BASES
656 #undef ENTRY
657       }
658     } else {
659       baseReg = MCOperand::CreateReg(0);
660     }
661 
662     // Check whether we are handling VSIB addressing mode for GATHER.
663     // If sibIndex was set to SIB_INDEX_NONE, index offset is 4 and
664     // we should use SIB_INDEX_XMM4|YMM4 for VSIB.
665     // I don't see a way to get the correct IndexReg in readSIB:
666     //   We can tell whether it is VSIB or SIB after instruction ID is decoded,
667     //   but instruction ID may not be decoded yet when calling readSIB.
668     uint32_t Opcode = mcInst.getOpcode();
669     bool IndexIs128 = (Opcode == X86::VGATHERDPDrm ||
670                        Opcode == X86::VGATHERDPDYrm ||
671                        Opcode == X86::VGATHERQPDrm ||
672                        Opcode == X86::VGATHERDPSrm ||
673                        Opcode == X86::VGATHERQPSrm ||
674                        Opcode == X86::VPGATHERDQrm ||
675                        Opcode == X86::VPGATHERDQYrm ||
676                        Opcode == X86::VPGATHERQQrm ||
677                        Opcode == X86::VPGATHERDDrm ||
678                        Opcode == X86::VPGATHERQDrm);
679     bool IndexIs256 = (Opcode == X86::VGATHERQPDYrm ||
680                        Opcode == X86::VGATHERDPSYrm ||
681                        Opcode == X86::VGATHERQPSYrm ||
682                        Opcode == X86::VGATHERDPDZrm ||
683                        Opcode == X86::VPGATHERDQZrm ||
684                        Opcode == X86::VPGATHERQQYrm ||
685                        Opcode == X86::VPGATHERDDYrm ||
686                        Opcode == X86::VPGATHERQDYrm);
687     bool IndexIs512 = (Opcode == X86::VGATHERQPDZrm ||
688                        Opcode == X86::VGATHERDPSZrm ||
689                        Opcode == X86::VGATHERQPSZrm ||
690                        Opcode == X86::VPGATHERQQZrm ||
691                        Opcode == X86::VPGATHERDDZrm ||
692                        Opcode == X86::VPGATHERQDZrm);
693     if (IndexIs128 || IndexIs256 || IndexIs512) {
694       unsigned IndexOffset = insn.sibIndex -
695                          (insn.addressSize == 8 ? SIB_INDEX_RAX:SIB_INDEX_EAX);
696       SIBIndex IndexBase = IndexIs512 ? SIB_INDEX_ZMM0 :
697                            IndexIs256 ? SIB_INDEX_YMM0 : SIB_INDEX_XMM0;
698       insn.sibIndex = (SIBIndex)(IndexBase +
699                            (insn.sibIndex == SIB_INDEX_NONE ? 4 : IndexOffset));
700     }
701 
702     if (insn.sibIndex != SIB_INDEX_NONE) {
703       switch (insn.sibIndex) {
704       default:
705         debug("Unexpected sibIndex");
706         return true;
707 #define ENTRY(x)                                          \
708       case SIB_INDEX_##x:                                 \
709         indexReg = MCOperand::CreateReg(X86::x); break;
710       EA_BASES_32BIT
711       EA_BASES_64BIT
712       REGS_XMM
713       REGS_YMM
714       REGS_ZMM
715 #undef ENTRY
716       }
717     } else {
718       indexReg = MCOperand::CreateReg(0);
719     }
720 
721     scaleAmount = MCOperand::CreateImm(insn.sibScale);
722   } else {
723     switch (insn.eaBase) {
724     case EA_BASE_NONE:
725       if (insn.eaDisplacement == EA_DISP_NONE) {
726         debug("EA_BASE_NONE and EA_DISP_NONE for ModR/M base");
727         return true;
728       }
729       if (insn.mode == MODE_64BIT){
730         pcrel = insn.startLocation +
731                 insn.displacementOffset + insn.displacementSize;
732         tryAddingPcLoadReferenceComment(insn.startLocation +
733                                         insn.displacementOffset,
734                                         insn.displacement + pcrel, Dis);
735         baseReg = MCOperand::CreateReg(X86::RIP); // Section 2.2.1.6
736       }
737       else
738         baseReg = MCOperand::CreateReg(0);
739 
740       indexReg = MCOperand::CreateReg(0);
741       break;
742     case EA_BASE_BX_SI:
743       baseReg = MCOperand::CreateReg(X86::BX);
744       indexReg = MCOperand::CreateReg(X86::SI);
745       break;
746     case EA_BASE_BX_DI:
747       baseReg = MCOperand::CreateReg(X86::BX);
748       indexReg = MCOperand::CreateReg(X86::DI);
749       break;
750     case EA_BASE_BP_SI:
751       baseReg = MCOperand::CreateReg(X86::BP);
752       indexReg = MCOperand::CreateReg(X86::SI);
753       break;
754     case EA_BASE_BP_DI:
755       baseReg = MCOperand::CreateReg(X86::BP);
756       indexReg = MCOperand::CreateReg(X86::DI);
757       break;
758     default:
759       indexReg = MCOperand::CreateReg(0);
760       switch (insn.eaBase) {
761       default:
762         debug("Unexpected eaBase");
763         return true;
764         // Here, we will use the fill-ins defined above.  However,
765         //   BX_SI, BX_DI, BP_SI, and BP_DI are all handled above and
766         //   sib and sib64 were handled in the top-level if, so they're only
767         //   placeholders to keep the compiler happy.
768 #define ENTRY(x)                                        \
769       case EA_BASE_##x:                                 \
770         baseReg = MCOperand::CreateReg(X86::x); break;
771       ALL_EA_BASES
772 #undef ENTRY
773 #define ENTRY(x) case EA_REG_##x:
774       ALL_REGS
775 #undef ENTRY
776         debug("A R/M memory operand may not be a register; "
777               "the base field must be a base.");
778         return true;
779       }
780     }
781 
782     scaleAmount = MCOperand::CreateImm(1);
783   }
784 
785   displacement = MCOperand::CreateImm(insn.displacement);
786 
787   segmentReg = MCOperand::CreateReg(segmentRegnums[insn.segmentOverride]);
788 
789   mcInst.addOperand(baseReg);
790   mcInst.addOperand(scaleAmount);
791   mcInst.addOperand(indexReg);
792   if(!tryAddingSymbolicOperand(insn.displacement + pcrel, false,
793                                insn.startLocation, insn.displacementOffset,
794                                insn.displacementSize, mcInst, Dis))
795     mcInst.addOperand(displacement);
796   mcInst.addOperand(segmentReg);
797   return false;
798 }
799 
800 /// translateRM - Translates an operand stored in the R/M (and possibly SIB)
801 ///   byte of an instruction to LLVM form, and appends it to an MCInst.
802 ///
803 /// @param mcInst       - The MCInst to append to.
804 /// @param operand      - The operand, as stored in the descriptor table.
805 /// @param insn         - The instruction to extract Mod, R/M, and SIB fields
806 ///                       from.
807 /// @return             - 0 on success; nonzero otherwise
translateRM(MCInst & mcInst,const OperandSpecifier & operand,InternalInstruction & insn,const MCDisassembler * Dis)808 static bool translateRM(MCInst &mcInst, const OperandSpecifier &operand,
809                         InternalInstruction &insn, const MCDisassembler *Dis) {
810   switch (operand.type) {
811   default:
812     debug("Unexpected type for a R/M operand");
813     return true;
814   case TYPE_R8:
815   case TYPE_R16:
816   case TYPE_R32:
817   case TYPE_R64:
818   case TYPE_Rv:
819   case TYPE_MM64:
820   case TYPE_XMM:
821   case TYPE_XMM32:
822   case TYPE_XMM64:
823   case TYPE_XMM128:
824   case TYPE_XMM256:
825   case TYPE_XMM512:
826   case TYPE_VK1:
827   case TYPE_VK8:
828   case TYPE_VK16:
829   case TYPE_DEBUGREG:
830   case TYPE_CONTROLREG:
831     return translateRMRegister(mcInst, insn);
832   case TYPE_M:
833   case TYPE_M8:
834   case TYPE_M16:
835   case TYPE_M32:
836   case TYPE_M64:
837   case TYPE_M128:
838   case TYPE_M256:
839   case TYPE_M512:
840   case TYPE_Mv:
841   case TYPE_M32FP:
842   case TYPE_M64FP:
843   case TYPE_M80FP:
844   case TYPE_M1616:
845   case TYPE_M1632:
846   case TYPE_M1664:
847   case TYPE_LEA:
848     return translateRMMemory(mcInst, insn, Dis);
849   }
850 }
851 
852 /// translateFPRegister - Translates a stack position on the FPU stack to its
853 ///   LLVM form, and appends it to an MCInst.
854 ///
855 /// @param mcInst       - The MCInst to append to.
856 /// @param stackPos     - The stack position to translate.
translateFPRegister(MCInst & mcInst,uint8_t stackPos)857 static void translateFPRegister(MCInst &mcInst,
858                                 uint8_t stackPos) {
859   mcInst.addOperand(MCOperand::CreateReg(X86::ST0 + stackPos));
860 }
861 
862 /// translateMaskRegister - Translates a 3-bit mask register number to
863 ///   LLVM form, and appends it to an MCInst.
864 ///
865 /// @param mcInst       - The MCInst to append to.
866 /// @param maskRegNum   - Number of mask register from 0 to 7.
867 /// @return             - false on success; true otherwise.
translateMaskRegister(MCInst & mcInst,uint8_t maskRegNum)868 static bool translateMaskRegister(MCInst &mcInst,
869                                 uint8_t maskRegNum) {
870   if (maskRegNum >= 8) {
871     debug("Invalid mask register number");
872     return true;
873   }
874 
875   mcInst.addOperand(MCOperand::CreateReg(X86::K0 + maskRegNum));
876   return false;
877 }
878 
879 /// translateOperand - Translates an operand stored in an internal instruction
880 ///   to LLVM's format and appends it to an MCInst.
881 ///
882 /// @param mcInst       - The MCInst to append to.
883 /// @param operand      - The operand, as stored in the descriptor table.
884 /// @param insn         - The internal instruction.
885 /// @return             - false on success; true otherwise.
translateOperand(MCInst & mcInst,const OperandSpecifier & operand,InternalInstruction & insn,const MCDisassembler * Dis)886 static bool translateOperand(MCInst &mcInst, const OperandSpecifier &operand,
887                              InternalInstruction &insn,
888                              const MCDisassembler *Dis) {
889   switch (operand.encoding) {
890   default:
891     debug("Unhandled operand encoding during translation");
892     return true;
893   case ENCODING_REG:
894     translateRegister(mcInst, insn.reg);
895     return false;
896   case ENCODING_WRITEMASK:
897     return translateMaskRegister(mcInst, insn.writemask);
898   CASE_ENCODING_RM:
899     return translateRM(mcInst, operand, insn, Dis);
900   case ENCODING_CB:
901   case ENCODING_CW:
902   case ENCODING_CD:
903   case ENCODING_CP:
904   case ENCODING_CO:
905   case ENCODING_CT:
906     debug("Translation of code offsets isn't supported.");
907     return true;
908   case ENCODING_IB:
909   case ENCODING_IW:
910   case ENCODING_ID:
911   case ENCODING_IO:
912   case ENCODING_Iv:
913   case ENCODING_Ia:
914     translateImmediate(mcInst,
915                        insn.immediates[insn.numImmediatesTranslated++],
916                        operand,
917                        insn,
918                        Dis);
919     return false;
920   case ENCODING_SI:
921     return translateSrcIndex(mcInst, insn);
922   case ENCODING_DI:
923     return translateDstIndex(mcInst, insn);
924   case ENCODING_RB:
925   case ENCODING_RW:
926   case ENCODING_RD:
927   case ENCODING_RO:
928   case ENCODING_Rv:
929     translateRegister(mcInst, insn.opcodeRegister);
930     return false;
931   case ENCODING_FP:
932     translateFPRegister(mcInst, insn.modRM & 7);
933     return false;
934   case ENCODING_VVVV:
935     translateRegister(mcInst, insn.vvvv);
936     return false;
937   case ENCODING_DUP:
938     return translateOperand(mcInst, insn.operands[operand.type - TYPE_DUP0],
939                             insn, Dis);
940   }
941 }
942 
943 /// translateInstruction - Translates an internal instruction and all its
944 ///   operands to an MCInst.
945 ///
946 /// @param mcInst       - The MCInst to populate with the instruction's data.
947 /// @param insn         - The internal instruction.
948 /// @return             - false on success; true otherwise.
translateInstruction(MCInst & mcInst,InternalInstruction & insn,const MCDisassembler * Dis)949 static bool translateInstruction(MCInst &mcInst,
950                                 InternalInstruction &insn,
951                                 const MCDisassembler *Dis) {
952   if (!insn.spec) {
953     debug("Instruction has no specification");
954     return true;
955   }
956 
957   mcInst.setOpcode(insn.instructionID);
958   // If when reading the prefix bytes we determined the overlapping 0xf2 or 0xf3
959   // prefix bytes should be disassembled as xrelease and xacquire then set the
960   // opcode to those instead of the rep and repne opcodes.
961   if (insn.xAcquireRelease) {
962     if(mcInst.getOpcode() == X86::REP_PREFIX)
963       mcInst.setOpcode(X86::XRELEASE_PREFIX);
964     else if(mcInst.getOpcode() == X86::REPNE_PREFIX)
965       mcInst.setOpcode(X86::XACQUIRE_PREFIX);
966   }
967 
968   insn.numImmediatesTranslated = 0;
969 
970   for (const auto &Op : insn.operands) {
971     if (Op.encoding != ENCODING_NONE) {
972       if (translateOperand(mcInst, Op, insn, Dis)) {
973         return true;
974       }
975     }
976   }
977 
978   return false;
979 }
980 
createX86Disassembler(const Target & T,const MCSubtargetInfo & STI,MCContext & Ctx)981 static MCDisassembler *createX86Disassembler(const Target &T,
982                                              const MCSubtargetInfo &STI,
983                                              MCContext &Ctx) {
984   std::unique_ptr<const MCInstrInfo> MII(T.createMCInstrInfo());
985   return new X86Disassembler::X86GenericDisassembler(STI, Ctx, std::move(MII));
986 }
987 
LLVMInitializeX86Disassembler()988 extern "C" void LLVMInitializeX86Disassembler() {
989   // Register the disassembler.
990   TargetRegistry::RegisterMCDisassembler(TheX86_32Target,
991                                          createX86Disassembler);
992   TargetRegistry::RegisterMCDisassembler(TheX86_64Target,
993                                          createX86Disassembler);
994 }
995