1 //===- Disassembler.cpp - Disassembler for hex strings --------------------===//
3 // The LLVM Compiler Infrastructure
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
8 //===----------------------------------------------------------------------===//
10 // This class implements the disassembler of strings of bytes written in
11 // hexadecimal, from standard input or from a file.
13 //===----------------------------------------------------------------------===//
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"
34 typedef std::vector<std::pair<unsigned char, const char*> > ByteArrayTy;
37 class VectorMemoryObject : public MemoryObject {
39 const ByteArrayTy &Bytes;
41 VectorMemoryObject(const ByteArrayTy &bytes) : Bytes(bytes) {}
43 uint64_t getBase() const { return 0; }
44 uint64_t getExtent() const { return Bytes.size(); }
46 int readByte(uint64_t Addr, uint8_t *Byte) const {
47 if (Addr >= getExtent())
49 *Byte = Bytes[Addr].first;
55 static 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);
61 // Disassemble it to strings.
65 for (Index = 0; Index < Bytes.size(); Index += Size) {
68 if (DisAsm.getInstruction(Inst, Size, memoryObject, Index,
69 /*REMOVE*/ nulls())) {
70 Printer.printInst(&Inst, Out);
73 SM.PrintMessage(SMLoc::getFromPointer(Bytes[Index].second),
74 "invalid instruction encoding", "warning");
76 Size = 1; // skip illegible bytes
83 static bool ByteArrayFromString(ByteArrayTy &ByteArray,
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);
93 // If this is the end of a line or start of a comment, remove the rest of
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.
101 Str = Str.substr(Str.find_first_of('\n'));
108 // Get the current token.
109 size_t Next = Str.find_first_of(" \t\n\r#");
110 StringRef Value = Str.substr(0, Next);
112 // Convert to a byte and add to the byte vector.
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'));
123 ByteArray.push_back(std::make_pair((unsigned char)ByteVal, Value.data()));
124 Str = Str.substr(Next);
130 int Disassembler::disassemble(const Target &T, const std::string &Triple,
131 MemoryBuffer &Buffer,
133 // Set up disassembler.
134 OwningPtr<const MCAsmInfo> AsmInfo(T.createAsmInfo(Triple));
137 errs() << "error: no assembly info for target " << Triple << "\n";
141 OwningPtr<const MCDisassembler> DisAsm(T.createMCDisassembler());
143 errs() << "error: no disassembler for target " << Triple << "\n";
147 int AsmPrinterVariant = AsmInfo->getAssemblerDialect();
148 OwningPtr<MCInstPrinter> IP(T.createMCInstPrinter(AsmPrinterVariant,
151 errs() << "error: no instruction printer for target " << Triple << '\n';
155 bool ErrorOccurred = false;
158 SM.AddNewSourceBuffer(&Buffer, SMLoc());
160 // Convert the input to a vector for disassembly.
161 ByteArrayTy ByteArray;
162 StringRef Str = Buffer.getBuffer();
164 ErrorOccurred |= ByteArrayFromString(ByteArray, Str, SM);
166 if (!ByteArray.empty())
167 ErrorOccurred |= PrintInsts(*DisAsm, *IP, ByteArray, SM, Out);
169 return ErrorOccurred;
172 static int byteArrayReader(uint8_t *B, uint64_t A, void *Arg) {
173 ByteArrayTy &ByteArray = *((ByteArrayTy*)Arg);
175 if (A >= ByteArray.size())
178 *B = ByteArray[A].first;
183 static int verboseEvaluator(uint64_t *V, unsigned R, void *Arg) {
184 EDDisassembler &disassembler = *(EDDisassembler *)((void **)Arg)[0];
185 raw_ostream &Out = *(raw_ostream *)((void **)Arg)[1];
187 if (const char *regName = disassembler.nameWithRegisterID(R))
188 Out << "[" << regName << "/" << R << "]";
190 if (disassembler.registerIsStackPointer(R))
192 if (disassembler.registerIsProgramCounter(R))
199 int Disassembler::disassembleEnhanced(const std::string &TS,
200 MemoryBuffer &Buffer,
202 ByteArrayTy ByteArray;
203 StringRef Str = Buffer.getBuffer();
206 SM.AddNewSourceBuffer(&Buffer, SMLoc());
208 if (ByteArrayFromString(ByteArray, Str, SM)) {
213 EDDisassembler::AssemblySyntax AS;
215 switch (T.getArch()) {
217 errs() << "error: no default assembly syntax for " << TS.c_str() << "\n";
221 AS = EDDisassembler::kEDAssemblySyntaxARMUAL;
225 AS = EDDisassembler::kEDAssemblySyntaxX86ATT;
229 EDDisassembler::initialize();
230 EDDisassembler *disassembler =
231 EDDisassembler::getDisassembler(TS.c_str(), AS);
233 if (disassembler == 0) {
234 errs() << "error: couldn't get disassembler for " << TS << '\n';
239 disassembler->createInst(byteArrayReader, 0, &ByteArray);
242 errs() << "error: Didn't get an instruction\n";
246 unsigned numTokens = inst->numTokens();
247 if ((int)numTokens < 0) {
248 errs() << "error: couldn't count the instruction's tokens\n";
252 for (unsigned tokenIndex = 0; tokenIndex != numTokens; ++tokenIndex) {
255 if (inst->getToken(token, tokenIndex)) {
256 errs() << "error: Couldn't get token\n";
261 if (token->getString(buf)) {
262 errs() << "error: Couldn't get string for token\n";
267 int operandIndex = token->operandID();
269 if (operandIndex >= 0)
270 Out << operandIndex << "-";
272 switch (token->type()) {
273 default: Out << "?"; break;
274 case EDToken::kTokenWhitespace: Out << "w"; break;
275 case EDToken::kTokenPunctuation: Out << "p"; break;
276 case EDToken::kTokenOpcode: Out << "o"; break;
277 case EDToken::kTokenLiteral: Out << "l"; break;
278 case EDToken::kTokenRegister: Out << "r"; break;
283 if (token->type() == EDToken::kTokenLiteral) {
285 if (token->literalSign())
287 uint64_t absoluteValue;
288 if (token->literalAbsoluteValue(absoluteValue)) {
289 errs() << "error: Couldn't get the value of a literal token\n";
292 Out << absoluteValue;
293 } else if (token->type() == EDToken::kTokenRegister) {
296 if (token->registerID(regID)) {
297 errs() << "error: Couldn't get the ID of a register token\n";
308 if (inst->isBranch())
313 unsigned numOperands = inst->numOperands();
315 if ((int)numOperands < 0) {
316 errs() << "error: Couldn't count operands\n";
320 for (unsigned operandIndex = 0; operandIndex != numOperands; ++operandIndex) {
321 Out << operandIndex << ":";
324 if (inst->getOperand(operand, operandIndex)) {
325 errs() << "error: couldn't get operand\n";
329 uint64_t evaluatedResult;
330 void *Arg[] = { disassembler, &Out };
331 evaluatedResult = operand->evaluate(evaluatedResult, verboseEvaluator, Arg);
332 Out << "=" << evaluatedResult << " ";