1//===-- llvm/MC/MCDisassembler.h - Disassembler 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#ifndef LLVM_MC_MCDISASSEMBLER_H 10#define LLVM_MC_MCDISASSEMBLER_H 11 12#include "llvm-c/Disassembler.h" 13#include "llvm/MC/MCRelocationInfo.h" 14#include "llvm/MC/MCSymbolizer.h" 15#include "llvm/Support/DataTypes.h" 16 17namespace llvm { 18 19class MCInst; 20class MCSubtargetInfo; 21class MemoryObject; 22class raw_ostream; 23class MCContext; 24 25/// MCDisassembler - Superclass for all disassemblers. Consumes a memory region 26/// and provides an array of assembly instructions. 27class MCDisassembler { 28public: 29 /// Ternary decode status. Most backends will just use Fail and 30 /// Success, however some have a concept of an instruction with 31 /// understandable semantics but which is architecturally 32 /// incorrect. An example of this is ARM UNPREDICTABLE instructions 33 /// which are disassemblable but cause undefined behaviour. 34 /// 35 /// Because it makes sense to disassemble these instructions, there 36 /// is a "soft fail" failure mode that indicates the MCInst& is 37 /// valid but architecturally incorrect. 38 /// 39 /// The enum numbers are deliberately chosen such that reduction 40 /// from Success->SoftFail ->Fail can be done with a simple 41 /// bitwise-AND: 42 /// 43 /// LEFT & TOP = | Success Unpredictable Fail 44 /// --------------+----------------------------------- 45 /// Success | Success Unpredictable Fail 46 /// Unpredictable | Unpredictable Unpredictable Fail 47 /// Fail | Fail Fail Fail 48 /// 49 /// An easy way of encoding this is as 0b11, 0b01, 0b00 for 50 /// Success, SoftFail, Fail respectively. 51 enum DecodeStatus { 52 Fail = 0, 53 SoftFail = 1, 54 Success = 3 55 }; 56 57 /// Constructor - Performs initial setup for the disassembler. 58 MCDisassembler(const MCSubtargetInfo &STI, MCContext &Ctx) 59 : Ctx(Ctx), STI(STI), Symbolizer(), CommentStream(nullptr) {} 60 61 virtual ~MCDisassembler(); 62 63 /// getInstruction - Returns the disassembly of a single instruction. 64 /// 65 /// @param instr - An MCInst to populate with the contents of the 66 /// instruction. 67 /// @param size - A value to populate with the size of the instruction, or 68 /// the number of bytes consumed while attempting to decode 69 /// an invalid instruction. 70 /// @param region - The memory object to use as a source for machine code. 71 /// @param address - The address, in the memory space of region, of the first 72 /// byte of the instruction. 73 /// @param vStream - The stream to print warnings and diagnostic messages on. 74 /// @param cStream - The stream to print comments and annotations on. 75 /// @return - MCDisassembler::Success if the instruction is valid, 76 /// MCDisassembler::SoftFail if the instruction was 77 /// disassemblable but invalid, 78 /// MCDisassembler::Fail if the instruction was invalid. 79 virtual DecodeStatus getInstruction(MCInst& instr, 80 uint64_t& size, 81 const MemoryObject ®ion, 82 uint64_t address, 83 raw_ostream &vStream, 84 raw_ostream &cStream) const = 0; 85private: 86 MCContext &Ctx; 87 88protected: 89 // Subtarget information, for instruction decoding predicates if required. 90 const MCSubtargetInfo &STI; 91 std::unique_ptr<MCSymbolizer> Symbolizer; 92 93public: 94 // Helpers around MCSymbolizer 95 bool tryAddingSymbolicOperand(MCInst &Inst, 96 int64_t Value, 97 uint64_t Address, bool IsBranch, 98 uint64_t Offset, uint64_t InstSize) const; 99 100 void tryAddingPcLoadReferenceComment(int64_t Value, uint64_t Address) const; 101 102 /// Set \p Symzer as the current symbolizer. 103 /// This takes ownership of \p Symzer, and deletes the previously set one. 104 void setSymbolizer(std::unique_ptr<MCSymbolizer> Symzer); 105 106 MCContext& getContext() const { return Ctx; } 107 108 const MCSubtargetInfo& getSubtargetInfo() const { return STI; } 109 110 // Marked mutable because we cache it inside the disassembler, rather than 111 // having to pass it around as an argument through all the autogenerated code. 112 mutable raw_ostream *CommentStream; 113}; 114 115} // namespace llvm 116 117#endif 118