1 //===-- lib/MC/Disassembler.cpp - Disassembler Public C Interface -*- C -*-===// 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 #include "Disassembler.h" 11 #include "llvm-c/Disassembler.h" 12 13 #include "llvm/MC/MCAsmInfo.h" 14 #include "llvm/MC/MCDisassembler.h" 15 #include "llvm/MC/MCInst.h" 16 #include "llvm/MC/MCInstPrinter.h" 17 #include "llvm/MC/MCContext.h" 18 #include "llvm/Target/TargetRegistry.h" 19 #include "llvm/Target/TargetAsmInfo.h" // FIXME. 20 #include "llvm/Target/TargetMachine.h" // FIXME. 21 #include "llvm/Target/TargetSelect.h" 22 #include "llvm/Support/MemoryObject.h" 23 24 namespace llvm { 25 class Target; 26 } // namespace llvm 27 using namespace llvm; 28 29 // LLVMCreateDisasm() creates a disassembler for the TripleName. Symbolic 30 // disassembly is supported by passing a block of information in the DisInfo 31 // parameter and specifying the TagType and callback functions as described in 32 // the header llvm-c/Disassembler.h . The pointer to the block and the 33 // functions can all be passed as NULL. If successful, this returns a 34 // disassembler context. If not, it returns NULL. 35 // 36 LLVMDisasmContextRef LLVMCreateDisasm(const char *TripleName, void *DisInfo, 37 int TagType, LLVMOpInfoCallback GetOpInfo, 38 LLVMSymbolLookupCallback SymbolLookUp) { 39 // Initialize targets and assembly printers/parsers. 40 llvm::InitializeAllTargetInfos(); 41 // FIXME: We shouldn't need to initialize the Target(Machine)s. 42 llvm::InitializeAllTargets(); 43 llvm::InitializeAllMCAsmInfos(); 44 llvm::InitializeAllAsmPrinters(); 45 llvm::InitializeAllAsmParsers(); 46 llvm::InitializeAllDisassemblers(); 47 48 // Get the target. 49 std::string Error; 50 const Target *TheTarget = TargetRegistry::lookupTarget(TripleName, Error); 51 assert(TheTarget && "Unable to create target!"); 52 53 // Get the assembler info needed to setup the MCContext. 54 const MCAsmInfo *MAI = TheTarget->createMCAsmInfo(TripleName); 55 assert(MAI && "Unable to create target asm info!"); 56 57 // Package up features to be passed to target/subtarget 58 std::string FeaturesStr; 59 std::string CPU; 60 61 // FIXME: We shouldn't need to do this (and link in codegen). 62 // When we split this out, we should do it in a way that makes 63 // it straightforward to switch subtargets on the fly. 64 TargetMachine *TM = TheTarget->createTargetMachine(TripleName, CPU, 65 FeaturesStr); 66 assert(TM && "Unable to create target machine!"); 67 68 // Get the target assembler info needed to setup the context. 69 const TargetAsmInfo *tai = new TargetAsmInfo(*TM); 70 assert(tai && "Unable to create target assembler!"); 71 72 // Set up the MCContext for creating symbols and MCExpr's. 73 MCContext *Ctx = new MCContext(*MAI, tai); 74 assert(Ctx && "Unable to create MCContext!"); 75 76 // Set up disassembler. 77 MCDisassembler *DisAsm = TheTarget->createMCDisassembler(); 78 assert(DisAsm && "Unable to create disassembler!"); 79 DisAsm->setupForSymbolicDisassembly(GetOpInfo, DisInfo, Ctx); 80 81 // Set up the instruction printer. 82 int AsmPrinterVariant = MAI->getAssemblerDialect(); 83 MCInstPrinter *IP = TheTarget->createMCInstPrinter(AsmPrinterVariant, 84 *MAI); 85 assert(IP && "Unable to create instruction printer!"); 86 87 LLVMDisasmContext *DC = new LLVMDisasmContext(TripleName, DisInfo, TagType, 88 GetOpInfo, SymbolLookUp, 89 TheTarget, MAI, TM, tai, Ctx, 90 DisAsm, IP); 91 assert(DC && "Allocation failure!"); 92 return DC; 93 } 94 95 // 96 // LLVMDisasmDispose() disposes of the disassembler specified by the context. 97 // 98 void LLVMDisasmDispose(LLVMDisasmContextRef DCR){ 99 LLVMDisasmContext *DC = (LLVMDisasmContext *)DCR; 100 delete DC; 101 } 102 103 namespace { 104 // 105 // The memory object created by LLVMDisasmInstruction(). 106 // 107 class DisasmMemoryObject : public MemoryObject { 108 uint8_t *Bytes; 109 uint64_t Size; 110 uint64_t BasePC; 111 public: 112 DisasmMemoryObject(uint8_t *bytes, uint64_t size, uint64_t basePC) : 113 Bytes(bytes), Size(size), BasePC(basePC) {} 114 115 uint64_t getBase() const { return BasePC; } 116 uint64_t getExtent() const { return Size; } 117 118 int readByte(uint64_t Addr, uint8_t *Byte) const { 119 if (Addr - BasePC >= Size) 120 return -1; 121 *Byte = Bytes[Addr - BasePC]; 122 return 0; 123 } 124 }; 125 } // end anonymous namespace 126 127 // 128 // LLVMDisasmInstruction() disassembles a single instruction using the 129 // disassembler context specified in the parameter DC. The bytes of the 130 // instruction are specified in the parameter Bytes, and contains at least 131 // BytesSize number of bytes. The instruction is at the address specified by 132 // the PC parameter. If a valid instruction can be disassembled its string is 133 // returned indirectly in OutString which whos size is specified in the 134 // parameter OutStringSize. This function returns the number of bytes in the 135 // instruction or zero if there was no valid instruction. If this function 136 // returns zero the caller will have to pick how many bytes they want to step 137 // over by printing a .byte, .long etc. to continue. 138 // 139 size_t LLVMDisasmInstruction(LLVMDisasmContextRef DCR, uint8_t *Bytes, 140 uint64_t BytesSize, uint64_t PC, char *OutString, 141 size_t OutStringSize){ 142 LLVMDisasmContext *DC = (LLVMDisasmContext *)DCR; 143 // Wrap the pointer to the Bytes, BytesSize and PC in a MemoryObject. 144 DisasmMemoryObject MemoryObject(Bytes, BytesSize, PC); 145 146 uint64_t Size; 147 MCInst Inst; 148 const MCDisassembler *DisAsm = DC->getDisAsm(); 149 MCInstPrinter *IP = DC->getIP(); 150 if (!DisAsm->getInstruction(Inst, Size, MemoryObject, PC, /*REMOVE*/ nulls())) 151 return 0; 152 153 SmallVector<char, 64> InsnStr; 154 raw_svector_ostream OS(InsnStr); 155 IP->printInst(&Inst, OS); 156 OS.flush(); 157 158 assert(OutStringSize != 0 && "Output buffer cannot be zero size"); 159 size_t OutputSize = std::min(OutStringSize-1, InsnStr.size()); 160 std::memcpy(OutString, InsnStr.data(), OutputSize); 161 OutString[OutputSize] = '\0'; // Terminate string. 162 163 return Size; 164 } 165