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 "llvm/ADT/Triple.h"
17#include "llvm/MC/MCAsmInfo.h"
18#include "llvm/MC/MCContext.h"
19#include "llvm/MC/MCDisassembler/MCDisassembler.h"
20#include "llvm/MC/MCInst.h"
21#include "llvm/MC/MCRegisterInfo.h"
22#include "llvm/MC/MCStreamer.h"
23#include "llvm/MC/MCSubtargetInfo.h"
24#include "llvm/Support/MemoryBuffer.h"
25#include "llvm/Support/SourceMgr.h"
26#include "llvm/Support/TargetRegistry.h"
27#include "llvm/Support/raw_ostream.h"
28
29using namespace llvm;
30
31typedef std::pair<std::vector<unsigned char>, std::vector<const char *>>
32    ByteArrayTy;
33
34static bool PrintInsts(const MCDisassembler &DisAsm,
35                       const ByteArrayTy &Bytes,
36                       SourceMgr &SM, raw_ostream &Out,
37                       MCStreamer &Streamer, bool InAtomicBlock,
38                       const MCSubtargetInfo &STI) {
39  ArrayRef<uint8_t> Data(Bytes.first.data(), Bytes.first.size());
40
41  // Disassemble it to strings.
42  uint64_t Size;
43  uint64_t Index;
44
45  for (Index = 0; Index < Bytes.first.size(); Index += Size) {
46    MCInst Inst;
47
48    MCDisassembler::DecodeStatus S;
49    S = DisAsm.getInstruction(Inst, Size, Data.slice(Index), Index,
50                              /*REMOVE*/ nulls(), nulls());
51    switch (S) {
52    case MCDisassembler::Fail:
53      SM.PrintMessage(SMLoc::getFromPointer(Bytes.second[Index]),
54                      SourceMgr::DK_Warning,
55                      "invalid instruction encoding");
56      // Don't try to resynchronise the stream in a block
57      if (InAtomicBlock)
58        return true;
59
60      if (Size == 0)
61        Size = 1; // skip illegible bytes
62
63      break;
64
65    case MCDisassembler::SoftFail:
66      SM.PrintMessage(SMLoc::getFromPointer(Bytes.second[Index]),
67                      SourceMgr::DK_Warning,
68                      "potentially undefined instruction encoding");
69      // Fall through
70
71    case MCDisassembler::Success:
72      Streamer.EmitInstruction(Inst, STI);
73      break;
74    }
75  }
76
77  return false;
78}
79
80static bool SkipToToken(StringRef &Str) {
81  for (;;) {
82    if (Str.empty())
83      return false;
84
85    // Strip horizontal whitespace and commas.
86    if (size_t Pos = Str.find_first_not_of(" \t\r\n,")) {
87      Str = Str.substr(Pos);
88      continue;
89    }
90
91    // If this is the start of a comment, remove the rest of the line.
92    if (Str[0] == '#') {
93        Str = Str.substr(Str.find_first_of('\n'));
94      continue;
95    }
96    return true;
97  }
98}
99
100
101static bool ByteArrayFromString(ByteArrayTy &ByteArray,
102                                StringRef &Str,
103                                SourceMgr &SM) {
104  while (SkipToToken(Str)) {
105    // Handled by higher level
106    if (Str[0] == '[' || Str[0] == ']')
107      return false;
108
109    // Get the current token.
110    size_t Next = Str.find_first_of(" \t\n\r,#[]");
111    StringRef Value = Str.substr(0, Next);
112
113    // Convert to a byte and add to the byte vector.
114    unsigned ByteVal;
115    if (Value.getAsInteger(0, ByteVal) || ByteVal > 255) {
116      // If we have an error, print it and skip to the end of line.
117      SM.PrintMessage(SMLoc::getFromPointer(Value.data()), SourceMgr::DK_Error,
118                      "invalid input token");
119      Str = Str.substr(Str.find('\n'));
120      ByteArray.first.clear();
121      ByteArray.second.clear();
122      continue;
123    }
124
125    ByteArray.first.push_back(ByteVal);
126    ByteArray.second.push_back(Value.data());
127    Str = Str.substr(Next);
128  }
129
130  return false;
131}
132
133int Disassembler::disassemble(const Target &T,
134                              const std::string &Triple,
135                              MCSubtargetInfo &STI,
136                              MCStreamer &Streamer,
137                              MemoryBuffer &Buffer,
138                              SourceMgr &SM,
139                              raw_ostream &Out) {
140
141  std::unique_ptr<const MCRegisterInfo> MRI(T.createMCRegInfo(Triple));
142  if (!MRI) {
143    errs() << "error: no register info for target " << Triple << "\n";
144    return -1;
145  }
146
147  std::unique_ptr<const MCAsmInfo> MAI(T.createMCAsmInfo(*MRI, Triple));
148  if (!MAI) {
149    errs() << "error: no assembly info for target " << Triple << "\n";
150    return -1;
151  }
152
153  // Set up the MCContext for creating symbols and MCExpr's.
154  MCContext Ctx(MAI.get(), MRI.get(), nullptr);
155
156  std::unique_ptr<const MCDisassembler> DisAsm(
157    T.createMCDisassembler(STI, Ctx));
158  if (!DisAsm) {
159    errs() << "error: no disassembler for target " << Triple << "\n";
160    return -1;
161  }
162
163  // Set up initial section manually here
164  Streamer.InitSections(false);
165
166  bool ErrorOccurred = false;
167
168  // Convert the input to a vector for disassembly.
169  ByteArrayTy ByteArray;
170  StringRef Str = Buffer.getBuffer();
171  bool InAtomicBlock = false;
172
173  while (SkipToToken(Str)) {
174    ByteArray.first.clear();
175    ByteArray.second.clear();
176
177    if (Str[0] == '[') {
178      if (InAtomicBlock) {
179        SM.PrintMessage(SMLoc::getFromPointer(Str.data()), SourceMgr::DK_Error,
180                        "nested atomic blocks make no sense");
181        ErrorOccurred = true;
182      }
183      InAtomicBlock = true;
184      Str = Str.drop_front();
185      continue;
186    } else if (Str[0] == ']') {
187      if (!InAtomicBlock) {
188        SM.PrintMessage(SMLoc::getFromPointer(Str.data()), SourceMgr::DK_Error,
189                        "attempt to close atomic block without opening");
190        ErrorOccurred = true;
191      }
192      InAtomicBlock = false;
193      Str = Str.drop_front();
194      continue;
195    }
196
197    // It's a real token, get the bytes and emit them
198    ErrorOccurred |= ByteArrayFromString(ByteArray, Str, SM);
199
200    if (!ByteArray.first.empty())
201      ErrorOccurred |= PrintInsts(*DisAsm, ByteArray, SM, Out, Streamer,
202                                  InAtomicBlock, STI);
203  }
204
205  if (InAtomicBlock) {
206    SM.PrintMessage(SMLoc::getFromPointer(Str.data()), SourceMgr::DK_Error,
207                    "unclosed atomic block");
208    ErrorOccurred = true;
209  }
210
211  return ErrorOccurred;
212}
213