1 //===- Disassembler.cpp - Disassembler for hex strings --------------------===// 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 class implements the disassembler of strings of bytes written in 11 // hexadecimal, from standard input or from a file. 12 // 13 //===----------------------------------------------------------------------===// 14 15 #include "Disassembler.h" 16 #include "../../lib/MC/MCDisassembler/EDDisassembler.h" 17 #include "../../lib/MC/MCDisassembler/EDInst.h" 18 #include "../../lib/MC/MCDisassembler/EDOperand.h" 19 #include "../../lib/MC/MCDisassembler/EDToken.h" 20 #include "llvm/MC/MCAsmInfo.h" 21 #include "llvm/MC/MCDisassembler.h" 22 #include "llvm/MC/MCInst.h" 23 #include "llvm/MC/MCInstPrinter.h" 24 #include "llvm/MC/MCRegisterInfo.h" 25 #include "llvm/MC/MCSubtargetInfo.h" 26 #include "llvm/ADT/OwningPtr.h" 27 #include "llvm/ADT/Triple.h" 28 #include "llvm/ADT/Twine.h" 29 #include "llvm/Support/MemoryBuffer.h" 30 #include "llvm/Support/MemoryObject.h" 31 #include "llvm/Support/SourceMgr.h" 32 #include "llvm/Support/TargetRegistry.h" 33 #include "llvm/Support/raw_ostream.h" 34 using namespace llvm; 35 36 typedef std::vector<std::pair<unsigned char, const char*> > ByteArrayTy; 37 38 namespace { 39 class VectorMemoryObject : public MemoryObject { 40 private: 41 const ByteArrayTy &Bytes; 42 public: 43 VectorMemoryObject(const ByteArrayTy &bytes) : Bytes(bytes) {} 44 45 uint64_t getBase() const { return 0; } 46 uint64_t getExtent() const { return Bytes.size(); } 47 48 int readByte(uint64_t Addr, uint8_t *Byte) const { 49 if (Addr >= getExtent()) 50 return -1; 51 *Byte = Bytes[Addr].first; 52 return 0; 53 } 54 }; 55 } 56 57 static bool PrintInsts(const MCDisassembler &DisAsm, 58 MCInstPrinter &Printer, const ByteArrayTy &Bytes, 59 SourceMgr &SM, raw_ostream &Out) { 60 // Wrap the vector in a MemoryObject. 61 VectorMemoryObject memoryObject(Bytes); 62 63 // Disassemble it to strings. 64 uint64_t Size; 65 uint64_t Index; 66 67 for (Index = 0; Index < Bytes.size(); Index += Size) { 68 MCInst Inst; 69 70 MCDisassembler::DecodeStatus S; 71 S = DisAsm.getInstruction(Inst, Size, memoryObject, Index, 72 /*REMOVE*/ nulls(), nulls()); 73 switch (S) { 74 case MCDisassembler::Fail: 75 SM.PrintMessage(SMLoc::getFromPointer(Bytes[Index].second), 76 SourceMgr::DK_Warning, 77 "invalid instruction encoding"); 78 if (Size == 0) 79 Size = 1; // skip illegible bytes 80 break; 81 82 case MCDisassembler::SoftFail: 83 SM.PrintMessage(SMLoc::getFromPointer(Bytes[Index].second), 84 SourceMgr::DK_Warning, 85 "potentially undefined instruction encoding"); 86 // Fall through 87 88 case MCDisassembler::Success: 89 Printer.printInst(&Inst, Out, ""); 90 Out << "\n"; 91 break; 92 } 93 } 94 95 return false; 96 } 97 98 static bool ByteArrayFromString(ByteArrayTy &ByteArray, 99 StringRef &Str, 100 SourceMgr &SM) { 101 while (!Str.empty()) { 102 // Strip horizontal whitespace. 103 if (size_t Pos = Str.find_first_not_of(" \t\r")) { 104 Str = Str.substr(Pos); 105 continue; 106 } 107 108 // If this is the end of a line or start of a comment, remove the rest of 109 // the line. 110 if (Str[0] == '\n' || Str[0] == '#') { 111 // Strip to the end of line if we already processed any bytes on this 112 // line. This strips the comment and/or the \n. 113 if (Str[0] == '\n') { 114 Str = Str.substr(1); 115 } else { 116 Str = Str.substr(Str.find_first_of('\n')); 117 if (!Str.empty()) 118 Str = Str.substr(1); 119 } 120 continue; 121 } 122 123 // Get the current token. 124 size_t Next = Str.find_first_of(" \t\n\r#"); 125 StringRef Value = Str.substr(0, Next); 126 127 // Convert to a byte and add to the byte vector. 128 unsigned ByteVal; 129 if (Value.getAsInteger(0, ByteVal) || ByteVal > 255) { 130 // If we have an error, print it and skip to the end of line. 131 SM.PrintMessage(SMLoc::getFromPointer(Value.data()), SourceMgr::DK_Error, 132 "invalid input token"); 133 Str = Str.substr(Str.find('\n')); 134 ByteArray.clear(); 135 continue; 136 } 137 138 ByteArray.push_back(std::make_pair((unsigned char)ByteVal, Value.data())); 139 Str = Str.substr(Next); 140 } 141 142 return false; 143 } 144 145 int Disassembler::disassemble(const Target &T, 146 const std::string &Triple, 147 const std::string &Cpu, 148 const std::string &FeaturesStr, 149 MemoryBuffer &Buffer, 150 raw_ostream &Out) { 151 // Set up disassembler. 152 OwningPtr<const MCAsmInfo> AsmInfo(T.createMCAsmInfo(Triple)); 153 154 if (!AsmInfo) { 155 errs() << "error: no assembly info for target " << Triple << "\n"; 156 return -1; 157 } 158 159 OwningPtr<const MCSubtargetInfo> STI(T.createMCSubtargetInfo(Triple, Cpu, FeaturesStr)); 160 if (!STI) { 161 errs() << "error: no subtarget info for target " << Triple << "\n"; 162 return -1; 163 } 164 165 OwningPtr<const MCDisassembler> DisAsm(T.createMCDisassembler(*STI)); 166 if (!DisAsm) { 167 errs() << "error: no disassembler for target " << Triple << "\n"; 168 return -1; 169 } 170 171 OwningPtr<const MCRegisterInfo> MRI(T.createMCRegInfo(Triple)); 172 if (!MRI) { 173 errs() << "error: no register info for target " << Triple << "\n"; 174 return -1; 175 } 176 177 int AsmPrinterVariant = AsmInfo->getAssemblerDialect(); 178 OwningPtr<MCInstPrinter> IP(T.createMCInstPrinter(AsmPrinterVariant, 179 *AsmInfo, *MRI, *STI)); 180 if (!IP) { 181 errs() << "error: no instruction printer for target " << Triple << '\n'; 182 return -1; 183 } 184 185 bool ErrorOccurred = false; 186 187 SourceMgr SM; 188 SM.AddNewSourceBuffer(&Buffer, SMLoc()); 189 190 // Convert the input to a vector for disassembly. 191 ByteArrayTy ByteArray; 192 StringRef Str = Buffer.getBuffer(); 193 194 ErrorOccurred |= ByteArrayFromString(ByteArray, Str, SM); 195 196 if (!ByteArray.empty()) 197 ErrorOccurred |= PrintInsts(*DisAsm, *IP, ByteArray, SM, Out); 198 199 return ErrorOccurred; 200 } 201 202 static int byteArrayReader(uint8_t *B, uint64_t A, void *Arg) { 203 ByteArrayTy &ByteArray = *((ByteArrayTy*)Arg); 204 205 if (A >= ByteArray.size()) 206 return -1; 207 208 *B = ByteArray[A].first; 209 210 return 0; 211 } 212 213 static int verboseEvaluator(uint64_t *V, unsigned R, void *Arg) { 214 EDDisassembler &disassembler = *(EDDisassembler *)((void **)Arg)[0]; 215 raw_ostream &Out = *(raw_ostream *)((void **)Arg)[1]; 216 217 if (const char *regName = disassembler.nameWithRegisterID(R)) 218 Out << "[" << regName << "/" << R << "]"; 219 220 if (disassembler.registerIsStackPointer(R)) 221 Out << "(sp)"; 222 if (disassembler.registerIsProgramCounter(R)) 223 Out << "(pc)"; 224 225 *V = 0; 226 return 0; 227 } 228 229 int Disassembler::disassembleEnhanced(const std::string &TS, 230 MemoryBuffer &Buffer, 231 raw_ostream &Out) { 232 ByteArrayTy ByteArray; 233 StringRef Str = Buffer.getBuffer(); 234 SourceMgr SM; 235 236 SM.AddNewSourceBuffer(&Buffer, SMLoc()); 237 238 if (ByteArrayFromString(ByteArray, Str, SM)) { 239 return -1; 240 } 241 242 Triple T(TS); 243 EDDisassembler::AssemblySyntax AS; 244 245 switch (T.getArch()) { 246 default: 247 errs() << "error: no default assembly syntax for " << TS.c_str() << "\n"; 248 return -1; 249 case Triple::arm: 250 case Triple::thumb: 251 AS = EDDisassembler::kEDAssemblySyntaxARMUAL; 252 break; 253 case Triple::x86: 254 case Triple::x86_64: 255 AS = EDDisassembler::kEDAssemblySyntaxX86ATT; 256 break; 257 } 258 259 OwningPtr<EDDisassembler> 260 disassembler(EDDisassembler::getDisassembler(TS.c_str(), AS)); 261 262 if (disassembler == 0) { 263 errs() << "error: couldn't get disassembler for " << TS << '\n'; 264 return -1; 265 } 266 267 while (ByteArray.size()) { 268 OwningPtr<EDInst> 269 inst(disassembler->createInst(byteArrayReader, 0, &ByteArray)); 270 271 if (inst == 0) { 272 errs() << "error: Didn't get an instruction\n"; 273 return -1; 274 } 275 276 ByteArray.erase (ByteArray.begin(), ByteArray.begin() + inst->byteSize()); 277 278 unsigned numTokens = inst->numTokens(); 279 if ((int)numTokens < 0) { 280 errs() << "error: couldn't count the instruction's tokens\n"; 281 return -1; 282 } 283 284 for (unsigned tokenIndex = 0; tokenIndex != numTokens; ++tokenIndex) { 285 EDToken *token; 286 287 if (inst->getToken(token, tokenIndex)) { 288 errs() << "error: Couldn't get token\n"; 289 return -1; 290 } 291 292 const char *buf; 293 if (token->getString(buf)) { 294 errs() << "error: Couldn't get string for token\n"; 295 return -1; 296 } 297 298 Out << '['; 299 int operandIndex = token->operandID(); 300 301 if (operandIndex >= 0) 302 Out << operandIndex << "-"; 303 304 switch (token->type()) { 305 case EDToken::kTokenWhitespace: Out << "w"; break; 306 case EDToken::kTokenPunctuation: Out << "p"; break; 307 case EDToken::kTokenOpcode: Out << "o"; break; 308 case EDToken::kTokenLiteral: Out << "l"; break; 309 case EDToken::kTokenRegister: Out << "r"; break; 310 } 311 312 Out << ":" << buf; 313 314 if (token->type() == EDToken::kTokenLiteral) { 315 Out << "="; 316 if (token->literalSign()) 317 Out << "-"; 318 uint64_t absoluteValue; 319 if (token->literalAbsoluteValue(absoluteValue)) { 320 errs() << "error: Couldn't get the value of a literal token\n"; 321 return -1; 322 } 323 Out << absoluteValue; 324 } else if (token->type() == EDToken::kTokenRegister) { 325 Out << "="; 326 unsigned regID; 327 if (token->registerID(regID)) { 328 errs() << "error: Couldn't get the ID of a register token\n"; 329 return -1; 330 } 331 Out << "r" << regID; 332 } 333 334 Out << "]"; 335 } 336 337 Out << " "; 338 339 if (inst->isBranch()) 340 Out << "<br> "; 341 if (inst->isMove()) 342 Out << "<mov> "; 343 344 unsigned numOperands = inst->numOperands(); 345 346 if ((int)numOperands < 0) { 347 errs() << "error: Couldn't count operands\n"; 348 return -1; 349 } 350 351 for (unsigned operandIndex = 0; operandIndex != numOperands; 352 ++operandIndex) { 353 Out << operandIndex << ":"; 354 355 EDOperand *operand; 356 if (inst->getOperand(operand, operandIndex)) { 357 errs() << "error: couldn't get operand\n"; 358 return -1; 359 } 360 361 uint64_t evaluatedResult; 362 void *Arg[] = { disassembler.get(), &Out }; 363 if (operand->evaluate(evaluatedResult, verboseEvaluator, Arg)) { 364 errs() << "error: Couldn't evaluate an operand\n"; 365 return -1; 366 } 367 Out << "=" << evaluatedResult << " "; 368 } 369 370 Out << '\n'; 371 } 372 373 return 0; 374 } 375