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