Disassembler.cpp revision 221337
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/Target/TargetRegistry.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/raw_ostream.h" 31#include "llvm/Support/SourceMgr.h" 32using namespace llvm; 33 34typedef std::vector<std::pair<unsigned char, const char*> > ByteArrayTy; 35 36namespace { 37class VectorMemoryObject : public MemoryObject { 38private: 39 const ByteArrayTy &Bytes; 40public: 41 VectorMemoryObject(const ByteArrayTy &bytes) : Bytes(bytes) {} 42 43 uint64_t getBase() const { return 0; } 44 uint64_t getExtent() const { return Bytes.size(); } 45 46 int readByte(uint64_t Addr, uint8_t *Byte) const { 47 if (Addr >= getExtent()) 48 return -1; 49 *Byte = Bytes[Addr].first; 50 return 0; 51 } 52}; 53} 54 55static bool PrintInsts(const MCDisassembler &DisAsm, 56 MCInstPrinter &Printer, const ByteArrayTy &Bytes, 57 SourceMgr &SM, raw_ostream &Out) { 58 // Wrap the vector in a MemoryObject. 59 VectorMemoryObject memoryObject(Bytes); 60 61 // Disassemble it to strings. 62 uint64_t Size; 63 uint64_t Index; 64 65 for (Index = 0; Index < Bytes.size(); Index += Size) { 66 MCInst Inst; 67 68 if (DisAsm.getInstruction(Inst, Size, memoryObject, Index, 69 /*REMOVE*/ nulls())) { 70 Printer.printInst(&Inst, Out); 71 Out << "\n"; 72 } else { 73 SM.PrintMessage(SMLoc::getFromPointer(Bytes[Index].second), 74 "invalid instruction encoding", "warning"); 75 if (Size == 0) 76 Size = 1; // skip illegible bytes 77 } 78 } 79 80 return false; 81} 82 83static bool ByteArrayFromString(ByteArrayTy &ByteArray, 84 StringRef &Str, 85 SourceMgr &SM) { 86 while (!Str.empty()) { 87 // Strip horizontal whitespace. 88 if (size_t Pos = Str.find_first_not_of(" \t\r")) { 89 Str = Str.substr(Pos); 90 continue; 91 } 92 93 // If this is the end of a line or start of a comment, remove the rest of 94 // the line. 95 if (Str[0] == '\n' || Str[0] == '#') { 96 // Strip to the end of line if we already processed any bytes on this 97 // line. This strips the comment and/or the \n. 98 if (Str[0] == '\n') { 99 Str = Str.substr(1); 100 } else { 101 Str = Str.substr(Str.find_first_of('\n')); 102 if (!Str.empty()) 103 Str = Str.substr(1); 104 } 105 continue; 106 } 107 108 // Get the current token. 109 size_t Next = Str.find_first_of(" \t\n\r#"); 110 StringRef Value = Str.substr(0, Next); 111 112 // Convert to a byte and add to the byte vector. 113 unsigned ByteVal; 114 if (Value.getAsInteger(0, ByteVal) || ByteVal > 255) { 115 // If we have an error, print it and skip to the end of line. 116 SM.PrintMessage(SMLoc::getFromPointer(Value.data()), 117 "invalid input token", "error"); 118 Str = Str.substr(Str.find('\n')); 119 ByteArray.clear(); 120 continue; 121 } 122 123 ByteArray.push_back(std::make_pair((unsigned char)ByteVal, Value.data())); 124 Str = Str.substr(Next); 125 } 126 127 return false; 128} 129 130int Disassembler::disassemble(const Target &T, TargetMachine &TM, 131 const std::string &Triple, 132 MemoryBuffer &Buffer, 133 raw_ostream &Out) { 134 // Set up disassembler. 135 OwningPtr<const MCAsmInfo> AsmInfo(T.createAsmInfo(Triple)); 136 137 if (!AsmInfo) { 138 errs() << "error: no assembly info for target " << Triple << "\n"; 139 return -1; 140 } 141 142 OwningPtr<const MCDisassembler> DisAsm(T.createMCDisassembler()); 143 if (!DisAsm) { 144 errs() << "error: no disassembler for target " << Triple << "\n"; 145 return -1; 146 } 147 148 int AsmPrinterVariant = AsmInfo->getAssemblerDialect(); 149 OwningPtr<MCInstPrinter> IP(T.createMCInstPrinter(TM, AsmPrinterVariant, 150 *AsmInfo)); 151 if (!IP) { 152 errs() << "error: no instruction printer for target " << Triple << '\n'; 153 return -1; 154 } 155 156 bool ErrorOccurred = false; 157 158 SourceMgr SM; 159 SM.AddNewSourceBuffer(&Buffer, SMLoc()); 160 161 // Convert the input to a vector for disassembly. 162 ByteArrayTy ByteArray; 163 StringRef Str = Buffer.getBuffer(); 164 165 ErrorOccurred |= ByteArrayFromString(ByteArray, Str, SM); 166 167 if (!ByteArray.empty()) 168 ErrorOccurred |= PrintInsts(*DisAsm, *IP, ByteArray, SM, Out); 169 170 return ErrorOccurred; 171} 172 173static int byteArrayReader(uint8_t *B, uint64_t A, void *Arg) { 174 ByteArrayTy &ByteArray = *((ByteArrayTy*)Arg); 175 176 if (A >= ByteArray.size()) 177 return -1; 178 179 *B = ByteArray[A].first; 180 181 return 0; 182} 183 184static int verboseEvaluator(uint64_t *V, unsigned R, void *Arg) { 185 EDDisassembler &disassembler = *(EDDisassembler *)((void **)Arg)[0]; 186 raw_ostream &Out = *(raw_ostream *)((void **)Arg)[1]; 187 188 if (const char *regName = disassembler.nameWithRegisterID(R)) 189 Out << "[" << regName << "/" << R << "]"; 190 191 if (disassembler.registerIsStackPointer(R)) 192 Out << "(sp)"; 193 if (disassembler.registerIsProgramCounter(R)) 194 Out << "(pc)"; 195 196 *V = 0; 197 return 0; 198} 199 200int Disassembler::disassembleEnhanced(const std::string &TS, 201 MemoryBuffer &Buffer, 202 raw_ostream &Out) { 203 ByteArrayTy ByteArray; 204 StringRef Str = Buffer.getBuffer(); 205 SourceMgr SM; 206 207 SM.AddNewSourceBuffer(&Buffer, SMLoc()); 208 209 if (ByteArrayFromString(ByteArray, Str, SM)) { 210 return -1; 211 } 212 213 Triple T(TS); 214 EDDisassembler::AssemblySyntax AS; 215 216 switch (T.getArch()) { 217 default: 218 errs() << "error: no default assembly syntax for " << TS.c_str() << "\n"; 219 return -1; 220 case Triple::arm: 221 case Triple::thumb: 222 AS = EDDisassembler::kEDAssemblySyntaxARMUAL; 223 break; 224 case Triple::x86: 225 case Triple::x86_64: 226 AS = EDDisassembler::kEDAssemblySyntaxX86ATT; 227 break; 228 } 229 230 EDDisassembler::initialize(); 231 OwningPtr<EDDisassembler> 232 disassembler(EDDisassembler::getDisassembler(TS.c_str(), AS)); 233 234 if (disassembler == 0) { 235 errs() << "error: couldn't get disassembler for " << TS << '\n'; 236 return -1; 237 } 238 239 while (ByteArray.size()) { 240 OwningPtr<EDInst> 241 inst(disassembler->createInst(byteArrayReader, 0, &ByteArray)); 242 243 if (inst == 0) { 244 errs() << "error: Didn't get an instruction\n"; 245 return -1; 246 } 247 248 ByteArray.erase (ByteArray.begin(), ByteArray.begin() + inst->byteSize()); 249 250 unsigned numTokens = inst->numTokens(); 251 if ((int)numTokens < 0) { 252 errs() << "error: couldn't count the instruction's tokens\n"; 253 return -1; 254 } 255 256 for (unsigned tokenIndex = 0; tokenIndex != numTokens; ++tokenIndex) { 257 EDToken *token; 258 259 if (inst->getToken(token, tokenIndex)) { 260 errs() << "error: Couldn't get token\n"; 261 return -1; 262 } 263 264 const char *buf; 265 if (token->getString(buf)) { 266 errs() << "error: Couldn't get string for token\n"; 267 return -1; 268 } 269 270 Out << '['; 271 int operandIndex = token->operandID(); 272 273 if (operandIndex >= 0) 274 Out << operandIndex << "-"; 275 276 switch (token->type()) { 277 default: Out << "?"; break; 278 case EDToken::kTokenWhitespace: Out << "w"; break; 279 case EDToken::kTokenPunctuation: Out << "p"; break; 280 case EDToken::kTokenOpcode: Out << "o"; break; 281 case EDToken::kTokenLiteral: Out << "l"; break; 282 case EDToken::kTokenRegister: Out << "r"; break; 283 } 284 285 Out << ":" << buf; 286 287 if (token->type() == EDToken::kTokenLiteral) { 288 Out << "="; 289 if (token->literalSign()) 290 Out << "-"; 291 uint64_t absoluteValue; 292 if (token->literalAbsoluteValue(absoluteValue)) { 293 errs() << "error: Couldn't get the value of a literal token\n"; 294 return -1; 295 } 296 Out << absoluteValue; 297 } else if (token->type() == EDToken::kTokenRegister) { 298 Out << "="; 299 unsigned regID; 300 if (token->registerID(regID)) { 301 errs() << "error: Couldn't get the ID of a register token\n"; 302 return -1; 303 } 304 Out << "r" << regID; 305 } 306 307 Out << "]"; 308 } 309 310 Out << " "; 311 312 if (inst->isBranch()) 313 Out << "<br> "; 314 if (inst->isMove()) 315 Out << "<mov> "; 316 317 unsigned numOperands = inst->numOperands(); 318 319 if ((int)numOperands < 0) { 320 errs() << "error: Couldn't count operands\n"; 321 return -1; 322 } 323 324 for (unsigned operandIndex = 0; operandIndex != numOperands; ++operandIndex) { 325 Out << operandIndex << ":"; 326 327 EDOperand *operand; 328 if (inst->getOperand(operand, operandIndex)) { 329 errs() << "error: couldn't get operand\n"; 330 return -1; 331 } 332 333 uint64_t evaluatedResult; 334 void *Arg[] = { disassembler.get(), &Out }; 335 if (operand->evaluate(evaluatedResult, verboseEvaluator, Arg)) { 336 errs() << "error: Couldn't evaluate an operand\n"; 337 return -1; 338 } 339 Out << "=" << evaluatedResult << " "; 340 } 341 342 Out << '\n'; 343 } 344 345 return 0; 346} 347 348