Disassembler.cpp revision 207618 
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
17#include "llvm/ADT/OwningPtr.h"
18#include "llvm/ADT/Triple.h"
19#include "llvm/MC/MCAsmInfo.h"
20#include "llvm/MC/MCDisassembler.h"
21#include "llvm/MC/MCInst.h"
22#include "llvm/MC/MCInstPrinter.h"
23#include "llvm/Target/TargetRegistry.h"
24#include "llvm/Support/MemoryBuffer.h"
25#include "llvm/Support/MemoryObject.h"
26#include "llvm/Support/raw_ostream.h"
27#include "llvm/Support/SourceMgr.h"
28
29#include "llvm-c/EnhancedDisassembly.h"
30
31using namespace llvm;
32
33typedef std::vector<std::pair<unsigned char, const char*> > ByteArrayTy;
34
35namespace {
36class VectorMemoryObject : public MemoryObject {
37private:
38  const ByteArrayTy &Bytes;
39public:
40  VectorMemoryObject(const ByteArrayTy &bytes) : Bytes(bytes) {}
41
42  uint64_t getBase() const { return 0; }
43  uint64_t getExtent() const { return Bytes.size(); }
44
45  int readByte(uint64_t Addr, uint8_t *Byte) const {
46    if (Addr > getExtent())
47      return -1;
48    *Byte = Bytes[Addr].first;
49    return 0;
50  }
51};
52}
53
54static bool PrintInsts(const MCDisassembler &DisAsm,
55                       MCInstPrinter &Printer, const ByteArrayTy &Bytes,
56                       SourceMgr &SM) {
57  // Wrap the vector in a MemoryObject.
58  VectorMemoryObject memoryObject(Bytes);
59
60  // Disassemble it to strings.
61  uint64_t Size;
62  uint64_t Index;
63
64  for (Index = 0; Index < Bytes.size(); Index += Size) {
65    MCInst Inst;
66
67    if (DisAsm.getInstruction(Inst, Size, memoryObject, Index,
68                               /*REMOVE*/ nulls())) {
69      Printer.printInst(&Inst, outs());
70      outs() << "\n";
71    } else {
72      SM.PrintMessage(SMLoc::getFromPointer(Bytes[Index].second),
73                      "invalid instruction encoding", "warning");
74      if (Size == 0)
75        Size = 1; // skip illegible bytes
76    }
77  }
78
79  return false;
80}
81
82static bool ByteArrayFromString(ByteArrayTy &ByteArray,
83                                StringRef &Str,
84                                SourceMgr &SM) {
85  while (!Str.empty()) {
86    // Strip horizontal whitespace.
87    if (size_t Pos = Str.find_first_not_of(" \t\r")) {
88      Str = Str.substr(Pos);
89      continue;
90    }
91
92    // If this is the end of a line or start of a comment, remove the rest of
93    // the line.
94    if (Str[0] == '\n' || Str[0] == '#') {
95      // Strip to the end of line if we already processed any bytes on this
96      // line.  This strips the comment and/or the \n.
97      if (Str[0] == '\n') {
98        Str = Str.substr(1);
99      } else {
100        Str = Str.substr(Str.find_first_of('\n'));
101        if (!Str.empty())
102          Str = Str.substr(1);
103      }
104      continue;
105    }
106
107    // Get the current token.
108    size_t Next = Str.find_first_of(" \t\n\r#");
109    StringRef Value = Str.substr(0, Next);
110
111    // Convert to a byte and add to the byte vector.
112    unsigned ByteVal;
113    if (Value.getAsInteger(0, ByteVal) || ByteVal > 255) {
114      // If we have an error, print it and skip to the end of line.
115      SM.PrintMessage(SMLoc::getFromPointer(Value.data()),
116                      "invalid input token", "error");
117      Str = Str.substr(Str.find('\n'));
118      ByteArray.clear();
119      continue;
120    }
121
122    ByteArray.push_back(std::make_pair((unsigned char)ByteVal, Value.data()));
123    Str = Str.substr(Next);
124  }
125
126  return false;
127}
128
129int Disassembler::disassemble(const Target &T, const std::string &Triple,
130                              MemoryBuffer &Buffer) {
131  // Set up disassembler.
132  OwningPtr<const MCAsmInfo> AsmInfo(T.createAsmInfo(Triple));
133
134  if (!AsmInfo) {
135    errs() << "error: no assembly info for target " << Triple << "\n";
136    return -1;
137  }
138
139  OwningPtr<const MCDisassembler> DisAsm(T.createMCDisassembler());
140  if (!DisAsm) {
141    errs() << "error: no disassembler for target " << Triple << "\n";
142    return -1;
143  }
144
145  int AsmPrinterVariant = AsmInfo->getAssemblerDialect();
146  OwningPtr<MCInstPrinter> IP(T.createMCInstPrinter(AsmPrinterVariant,
147                                                    *AsmInfo));
148  if (!IP) {
149    errs() << "error: no instruction printer for target " << Triple << '\n';
150    return -1;
151  }
152
153  bool ErrorOccurred = false;
154
155  SourceMgr SM;
156  SM.AddNewSourceBuffer(&Buffer, SMLoc());
157
158  // Convert the input to a vector for disassembly.
159  ByteArrayTy ByteArray;
160  StringRef Str = Buffer.getBuffer();
161
162  ErrorOccurred |= ByteArrayFromString(ByteArray, Str, SM);
163
164  if (!ByteArray.empty())
165    ErrorOccurred |= PrintInsts(*DisAsm, *IP, ByteArray, SM);
166
167  return ErrorOccurred;
168}
169
170static int byteArrayReader(uint8_t *B, uint64_t A, void *Arg) {
171  ByteArrayTy &ByteArray = *((ByteArrayTy*)Arg);
172
173  if (A >= ByteArray.size())
174    return -1;
175
176  *B = ByteArray[A].first;
177
178  return 0;
179}
180
181static int verboseEvaluator(uint64_t *V, unsigned R, void *Arg) {
182  EDDisassemblerRef &disassembler = *((EDDisassemblerRef*)Arg);
183
184  const char *regName;
185
186  if (!EDGetRegisterName(&regName,
187                        disassembler,
188                        R))
189    outs() << "[" << regName << "/" << R << "]";
190  if (EDRegisterIsStackPointer(disassembler, R))
191    outs() << "(sp)";
192  if (EDRegisterIsProgramCounter(disassembler, R))
193    outs() << "(pc)";
194
195  *V = 0;
196
197  return 0;
198}
199
200int Disassembler::disassembleEnhanced(const std::string &TS,
201                                      MemoryBuffer &Buffer) {
202  ByteArrayTy ByteArray;
203  StringRef Str = Buffer.getBuffer();
204  SourceMgr SM;
205
206  SM.AddNewSourceBuffer(&Buffer, SMLoc());
207
208  if (ByteArrayFromString(ByteArray, Str, SM)) {
209    return -1;
210  }
211
212  EDDisassemblerRef disassembler;
213
214  Triple T(TS);
215  EDAssemblySyntax_t AS;
216
217  switch (T.getArch()) {
218  default:
219    errs() << "error: no default assembly syntax for " << TS.c_str() << "\n";
220    return -1;
221  case Triple::arm:
222  case Triple::thumb:
223    AS = kEDAssemblySyntaxARMUAL;
224    break;
225  case Triple::x86:
226  case Triple::x86_64:
227    AS = kEDAssemblySyntaxX86ATT;
228    break;
229  }
230
231  if (EDGetDisassembler(&disassembler,
232                        TS.c_str(),
233                        AS)) {
234    errs() << "error: couldn't get disassembler for " << TS.c_str() << "\n";
235    return -1;
236  }
237
238  EDInstRef inst;
239
240  if (EDCreateInsts(&inst, 1, disassembler, byteArrayReader, 0,&ByteArray)
241      != 1) {
242    errs() << "error: Didn't get an instruction\n";
243    return -1;
244  }
245
246  int numTokens = EDNumTokens(inst);
247
248  if (numTokens < 0) {
249    errs() << "error: Couldn't count the instruction's tokens\n";
250    return -1;
251  }
252
253  int tokenIndex;
254
255  for (tokenIndex = 0; tokenIndex < numTokens; ++tokenIndex) {
256    EDTokenRef token;
257
258    if (EDGetToken(&token, inst, tokenIndex)) {
259      errs() << "error: Couldn't get token\n";
260      return -1;
261    }
262
263    const char *buf;
264
265    if (EDGetTokenString(&buf, token)) {
266      errs() << "error: Couldn't get string for token\n";
267      return -1;
268    }
269
270    outs() << "[";
271
272    int operandIndex = EDOperandIndexForToken(token);
273
274    if (operandIndex >= 0)
275      outs() << operandIndex << "-";
276
277    if (EDTokenIsWhitespace(token)) {
278      outs() << "w";
279    } else if (EDTokenIsPunctuation(token)) {
280      outs() << "p";
281    } else if (EDTokenIsOpcode(token)) {
282      outs() << "o";
283    } else if (EDTokenIsLiteral(token)) {
284      outs() << "l";
285    } else if (EDTokenIsRegister(token)) {
286      outs() << "r";
287    } else {
288      outs() << "?";
289    }
290
291    outs() << ":" << buf;
292
293    if (EDTokenIsLiteral(token)) {
294      outs() << "=";
295      if (EDTokenIsNegativeLiteral(token))
296        outs() << "-";
297      uint64_t absoluteValue;
298      if (EDLiteralTokenAbsoluteValue(&absoluteValue, token)) {
299        errs() << "error: Couldn't get the value of a literal token\n";
300        return -1;
301      }
302      outs() << absoluteValue;
303    } else if (EDTokenIsRegister(token)) {
304      outs() << "=";
305      unsigned regID;
306      if (EDRegisterTokenValue(&regID, token)) {
307        errs() << "error: Couldn't get the ID of a register token\n";
308        return -1;
309      }
310      outs() << "r" << regID;
311    }
312
313    outs() << "]";
314  }
315
316  outs() << " ";
317
318  int numOperands = EDNumOperands(inst);
319
320  if (numOperands < 0) {
321    errs() << "error: Couldn't count operands\n";
322    return -1;
323  }
324
325  int operandIndex;
326
327  for (operandIndex = 0; operandIndex < numOperands; ++operandIndex) {
328    outs() << operandIndex << ":";
329
330    EDOperandRef operand;
331
332    if (EDGetOperand(&operand,
333                     inst,
334                     operandIndex)) {
335      errs() << "error: Couldn't get operand\n";
336      return -1;
337    }
338
339    uint64_t evaluatedResult;
340
341    EDEvaluateOperand(&evaluatedResult,
342                      operand,
343                      verboseEvaluator,
344                      &disassembler);
345
346    outs() << " ";
347  }
348
349  outs() << "\n";
350
351  return 0;
352}
353
354