1 //===-- lib/MC/Disassembler.cpp - Disassembler Public C Interface ---------===//
3 // The LLVM Compiler Infrastructure
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
8 //===----------------------------------------------------------------------===//
10 #include "Disassembler.h"
11 #include "llvm-c/Disassembler.h"
13 #include "llvm/MC/MCAsmInfo.h"
14 #include "llvm/MC/MCContext.h"
15 #include "llvm/MC/MCDisassembler.h"
16 #include "llvm/MC/MCInst.h"
17 #include "llvm/MC/MCInstPrinter.h"
18 #include "llvm/MC/MCRegisterInfo.h"
19 #include "llvm/Support/MemoryObject.h"
20 #include "llvm/Support/TargetRegistry.h"
21 #include "llvm/Support/TargetSelect.h"
22 #include "llvm/Support/ErrorHandling.h"
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.
36 LLVMDisasmContextRef LLVMCreateDisasm(const char *TripleName, void *DisInfo,
37 int TagType, LLVMOpInfoCallback GetOpInfo,
38 LLVMSymbolLookupCallback SymbolLookUp) {
39 // Initialize targets and assembly printers/parsers.
40 // FIXME: Clients are responsible for initializing the targets. And this
41 // would be done by calling routines in "llvm-c/Target.h" which are static
42 // line functions. But the current use of LLVMCreateDisasm() is to dynamically
43 // load libLTO with dlopen() and then lookup the symbols using dlsym().
44 // And since these initialize routines are static that does not work which
45 // is why the call to them in this 'C' library API was added back.
46 llvm::InitializeAllTargetInfos();
47 llvm::InitializeAllTargetMCs();
48 llvm::InitializeAllAsmParsers();
49 llvm::InitializeAllDisassemblers();
53 const Target *TheTarget = TargetRegistry::lookupTarget(TripleName, Error);
54 assert(TheTarget && "Unable to create target!");
56 // Get the assembler info needed to setup the MCContext.
57 const MCAsmInfo *MAI = TheTarget->createMCAsmInfo(TripleName);
58 assert(MAI && "Unable to create target asm info!");
60 const MCRegisterInfo *MRI = TheTarget->createMCRegInfo(TripleName);
61 assert(MRI && "Unable to create target register info!");
63 // Package up features to be passed to target/subtarget
64 std::string FeaturesStr;
67 const MCSubtargetInfo *STI = TheTarget->createMCSubtargetInfo(TripleName, CPU,
69 assert(STI && "Unable to create subtarget info!");
71 // Set up the MCContext for creating symbols and MCExpr's.
72 MCContext *Ctx = new MCContext(*MAI, *MRI, 0);
73 assert(Ctx && "Unable to create MCContext!");
75 // Set up disassembler.
76 MCDisassembler *DisAsm = TheTarget->createMCDisassembler(*STI);
77 assert(DisAsm && "Unable to create disassembler!");
78 DisAsm->setupForSymbolicDisassembly(GetOpInfo, SymbolLookUp, DisInfo, Ctx);
80 // Set up the instruction printer.
81 int AsmPrinterVariant = MAI->getAssemblerDialect();
82 MCInstPrinter *IP = TheTarget->createMCInstPrinter(AsmPrinterVariant,
84 assert(IP && "Unable to create instruction printer!");
86 LLVMDisasmContext *DC = new LLVMDisasmContext(TripleName, DisInfo, TagType,
87 GetOpInfo, SymbolLookUp,
90 assert(DC && "Allocation failure!");
96 // LLVMDisasmDispose() disposes of the disassembler specified by the context.
98 void LLVMDisasmDispose(LLVMDisasmContextRef DCR){
99 LLVMDisasmContext *DC = (LLVMDisasmContext *)DCR;
105 // The memory object created by LLVMDisasmInstruction().
107 class DisasmMemoryObject : public MemoryObject {
112 DisasmMemoryObject(uint8_t *bytes, uint64_t size, uint64_t basePC) :
113 Bytes(bytes), Size(size), BasePC(basePC) {}
115 uint64_t getBase() const { return BasePC; }
116 uint64_t getExtent() { return Size; }
118 int readByte(uint64_t Addr, uint8_t *Byte) {
119 if (Addr - BasePC >= Size)
121 *Byte = Bytes[Addr - BasePC];
125 } // end anonymous namespace
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.
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);
148 const MCDisassembler *DisAsm = DC->getDisAsm();
149 MCInstPrinter *IP = DC->getIP();
150 MCDisassembler::DecodeStatus S;
151 S = DisAsm->getInstruction(Inst, Size, MemoryObject, PC,
152 /*REMOVE*/ nulls(), DC->CommentStream);
154 case MCDisassembler::Fail:
155 case MCDisassembler::SoftFail:
156 // FIXME: Do something different for soft failure modes?
159 case MCDisassembler::Success: {
160 DC->CommentStream.flush();
161 StringRef Comments = DC->CommentsToEmit.str();
163 SmallVector<char, 64> InsnStr;
164 raw_svector_ostream OS(InsnStr);
165 IP->printInst(&Inst, OS, Comments);
168 // Tell the comment stream that the vector changed underneath it.
169 DC->CommentsToEmit.clear();
170 DC->CommentStream.resync();
172 assert(OutStringSize != 0 && "Output buffer cannot be zero size");
173 size_t OutputSize = std::min(OutStringSize-1, InsnStr.size());
174 std::memcpy(OutString, InsnStr.data(), OutputSize);
175 OutString[OutputSize] = '\0'; // Terminate string.
180 llvm_unreachable("Invalid DecodeStatus!");