1 //===- Disassembler.cpp - Disassembler for hex strings --------------------===// 2 // 3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. 4 // See https://llvm.org/LICENSE.txt for license information. 5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception 6 // 7 //===----------------------------------------------------------------------===// 8 // 9 // This class implements the disassembler of strings of bytes written in 10 // hexadecimal, from standard input or from a file. 11 // 12 //===----------------------------------------------------------------------===// 13 14 #include "Disassembler.h" 15 #include "llvm/ADT/Triple.h" 16 #include "llvm/MC/MCAsmInfo.h" 17 #include "llvm/MC/MCContext.h" 18 #include "llvm/MC/MCDisassembler/MCDisassembler.h" 19 #include "llvm/MC/MCInst.h" 20 #include "llvm/MC/MCObjectFileInfo.h" 21 #include "llvm/MC/MCRegisterInfo.h" 22 #include "llvm/MC/MCStreamer.h" 23 #include "llvm/MC/MCSubtargetInfo.h" 24 #include "llvm/MC/TargetRegistry.h" 25 #include "llvm/Support/MemoryBuffer.h" 26 #include "llvm/Support/SourceMgr.h" 27 #include "llvm/Support/raw_ostream.h" 28 29 using namespace llvm; 30 31 typedef std::pair<std::vector<unsigned char>, std::vector<const char *>> 32 ByteArrayTy; 33 34 static 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, nulls()); 50 switch (S) { 51 case MCDisassembler::Fail: 52 SM.PrintMessage(SMLoc::getFromPointer(Bytes.second[Index]), 53 SourceMgr::DK_Warning, 54 "invalid instruction encoding"); 55 // Don't try to resynchronise the stream in a block 56 if (InAtomicBlock) 57 return true; 58 59 if (Size == 0) 60 Size = 1; // skip illegible bytes 61 62 break; 63 64 case MCDisassembler::SoftFail: 65 SM.PrintMessage(SMLoc::getFromPointer(Bytes.second[Index]), 66 SourceMgr::DK_Warning, 67 "potentially undefined instruction encoding"); 68 [[fallthrough]]; 69 70 case MCDisassembler::Success: 71 Streamer.emitInstruction(Inst, STI); 72 break; 73 } 74 } 75 76 return false; 77 } 78 79 static bool SkipToToken(StringRef &Str) { 80 for (;;) { 81 if (Str.empty()) 82 return false; 83 84 // Strip horizontal whitespace and commas. 85 if (size_t Pos = Str.find_first_not_of(" \t\r\n,")) { 86 Str = Str.substr(Pos); 87 continue; 88 } 89 90 // If this is the start of a comment, remove the rest of the line. 91 if (Str[0] == '#') { 92 Str = Str.substr(Str.find_first_of('\n')); 93 continue; 94 } 95 return true; 96 } 97 } 98 99 100 static bool ByteArrayFromString(ByteArrayTy &ByteArray, 101 StringRef &Str, 102 SourceMgr &SM) { 103 while (SkipToToken(Str)) { 104 // Handled by higher level 105 if (Str[0] == '[' || Str[0] == ']') 106 return false; 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()), SourceMgr::DK_Error, 117 "invalid input token"); 118 Str = Str.substr(Str.find('\n')); 119 ByteArray.first.clear(); 120 ByteArray.second.clear(); 121 continue; 122 } 123 124 ByteArray.first.push_back(ByteVal); 125 ByteArray.second.push_back(Value.data()); 126 Str = Str.substr(Next); 127 } 128 129 return false; 130 } 131 132 int Disassembler::disassemble(const Target &T, const std::string &Triple, 133 MCSubtargetInfo &STI, MCStreamer &Streamer, 134 MemoryBuffer &Buffer, SourceMgr &SM, 135 MCContext &Ctx, raw_ostream &Out, 136 const MCTargetOptions &MCOptions) { 137 138 std::unique_ptr<const MCRegisterInfo> MRI(T.createMCRegInfo(Triple)); 139 if (!MRI) { 140 errs() << "error: no register info for target " << Triple << "\n"; 141 return -1; 142 } 143 144 std::unique_ptr<const MCAsmInfo> MAI( 145 T.createMCAsmInfo(*MRI, Triple, MCOptions)); 146 if (!MAI) { 147 errs() << "error: no assembly info for target " << Triple << "\n"; 148 return -1; 149 } 150 151 std::unique_ptr<const MCDisassembler> DisAsm( 152 T.createMCDisassembler(STI, Ctx)); 153 if (!DisAsm) { 154 errs() << "error: no disassembler for target " << Triple << "\n"; 155 return -1; 156 } 157 158 // Set up initial section manually here 159 Streamer.initSections(false, STI); 160 161 bool ErrorOccurred = false; 162 163 // Convert the input to a vector for disassembly. 164 ByteArrayTy ByteArray; 165 StringRef Str = Buffer.getBuffer(); 166 bool InAtomicBlock = false; 167 168 while (SkipToToken(Str)) { 169 ByteArray.first.clear(); 170 ByteArray.second.clear(); 171 172 if (Str[0] == '[') { 173 if (InAtomicBlock) { 174 SM.PrintMessage(SMLoc::getFromPointer(Str.data()), SourceMgr::DK_Error, 175 "nested atomic blocks make no sense"); 176 ErrorOccurred = true; 177 } 178 InAtomicBlock = true; 179 Str = Str.drop_front(); 180 continue; 181 } else if (Str[0] == ']') { 182 if (!InAtomicBlock) { 183 SM.PrintMessage(SMLoc::getFromPointer(Str.data()), SourceMgr::DK_Error, 184 "attempt to close atomic block without opening"); 185 ErrorOccurred = true; 186 } 187 InAtomicBlock = false; 188 Str = Str.drop_front(); 189 continue; 190 } 191 192 // It's a real token, get the bytes and emit them 193 ErrorOccurred |= ByteArrayFromString(ByteArray, Str, SM); 194 195 if (!ByteArray.first.empty()) 196 ErrorOccurred |= PrintInsts(*DisAsm, ByteArray, SM, Out, Streamer, 197 InAtomicBlock, STI); 198 } 199 200 if (InAtomicBlock) { 201 SM.PrintMessage(SMLoc::getFromPointer(Str.data()), SourceMgr::DK_Error, 202 "unclosed atomic block"); 203 ErrorOccurred = true; 204 } 205 206 return ErrorOccurred; 207 } 208