Disassembler.cpp revision 98c5ddabca1debf935a07d14d0cbc9732374bdb8 
1//===-- lib/MC/Disassembler.cpp - Disassembler Public C Interface ---------===//
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/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
23namespace llvm {
24class Target;
25} // namespace llvm
26using namespace llvm;
27
28// LLVMCreateDisasm() creates a disassembler for the TripleName.  Symbolic
29// disassembly is supported by passing a block of information in the DisInfo
30// parameter and specifying the TagType and callback functions as described in
31// the header llvm-c/Disassembler.h .  The pointer to the block and the
32// functions can all be passed as NULL.  If successful, this returns a
33// disassembler context.  If not, it returns NULL.
34//
35LLVMDisasmContextRef LLVMCreateDisasm(const char *TripleName, void *DisInfo,
36                                      int TagType, LLVMOpInfoCallback GetOpInfo,
37                                      LLVMSymbolLookupCallback SymbolLookUp) {
38  // Initialize targets and assembly printers/parsers.
39  llvm::InitializeAllTargetInfos();
40  llvm::InitializeAllTargetMCs();
41  llvm::InitializeAllAsmParsers();
42  llvm::InitializeAllDisassemblers();
43
44  // Get the target.
45  std::string Error;
46  const Target *TheTarget = TargetRegistry::lookupTarget(TripleName, Error);
47  assert(TheTarget && "Unable to create target!");
48
49  // Get the assembler info needed to setup the MCContext.
50  const MCAsmInfo *MAI = TheTarget->createMCAsmInfo(TripleName);
51  assert(MAI && "Unable to create target asm info!");
52
53  const MCRegisterInfo *MRI = TheTarget->createMCRegInfo(TripleName);
54  assert(MRI && "Unable to create target register info!");
55
56  // Package up features to be passed to target/subtarget
57  std::string FeaturesStr;
58  std::string CPU;
59
60  const MCSubtargetInfo *STI = TheTarget->createMCSubtargetInfo(TripleName, CPU,
61                                                                FeaturesStr);
62  assert(STI && "Unable to create subtarget info!");
63
64  // Set up the MCContext for creating symbols and MCExpr's.
65  MCContext *Ctx = new MCContext(*MAI, *MRI, 0);
66  assert(Ctx && "Unable to create MCContext!");
67
68  // Set up disassembler.
69  MCDisassembler *DisAsm = TheTarget->createMCDisassembler(*STI);
70  assert(DisAsm && "Unable to create disassembler!");
71  DisAsm->setupForSymbolicDisassembly(GetOpInfo, DisInfo, Ctx);
72
73  // Set up the instruction printer.
74  int AsmPrinterVariant = MAI->getAssemblerDialect();
75  MCInstPrinter *IP = TheTarget->createMCInstPrinter(AsmPrinterVariant,
76                                                     *MAI, *STI);
77  assert(IP && "Unable to create instruction printer!");
78
79  LLVMDisasmContext *DC = new LLVMDisasmContext(TripleName, DisInfo, TagType,
80                                                GetOpInfo, SymbolLookUp,
81                                                TheTarget, MAI, MRI,
82                                                Ctx, DisAsm, IP);
83  assert(DC && "Allocation failure!");
84
85  IP->setCommentStream(DC->CommentStream);
86
87  return DC;
88}
89
90//
91// LLVMDisasmDispose() disposes of the disassembler specified by the context.
92//
93void LLVMDisasmDispose(LLVMDisasmContextRef DCR){
94  LLVMDisasmContext *DC = (LLVMDisasmContext *)DCR;
95  delete DC;
96}
97
98namespace {
99//
100// The memory object created by LLVMDisasmInstruction().
101//
102class DisasmMemoryObject : public MemoryObject {
103  uint8_t *Bytes;
104  uint64_t Size;
105  uint64_t BasePC;
106public:
107  DisasmMemoryObject(uint8_t *bytes, uint64_t size, uint64_t basePC) :
108                     Bytes(bytes), Size(size), BasePC(basePC) {}
109
110  uint64_t getBase() const { return BasePC; }
111  uint64_t getExtent() const { return Size; }
112
113  int readByte(uint64_t Addr, uint8_t *Byte) const {
114    if (Addr - BasePC >= Size)
115      return -1;
116    *Byte = Bytes[Addr - BasePC];
117    return 0;
118  }
119};
120} // end anonymous namespace
121
122//
123// LLVMDisasmInstruction() disassembles a single instruction using the
124// disassembler context specified in the parameter DC.  The bytes of the
125// instruction are specified in the parameter Bytes, and contains at least
126// BytesSize number of bytes.  The instruction is at the address specified by
127// the PC parameter.  If a valid instruction can be disassembled its string is
128// returned indirectly in OutString which whos size is specified in the
129// parameter OutStringSize.  This function returns the number of bytes in the
130// instruction or zero if there was no valid instruction.  If this function
131// returns zero the caller will have to pick how many bytes they want to step
132// over by printing a .byte, .long etc. to continue.
133//
134size_t LLVMDisasmInstruction(LLVMDisasmContextRef DCR, uint8_t *Bytes,
135                             uint64_t BytesSize, uint64_t PC, char *OutString,
136                             size_t OutStringSize){
137  LLVMDisasmContext *DC = (LLVMDisasmContext *)DCR;
138  // Wrap the pointer to the Bytes, BytesSize and PC in a MemoryObject.
139  DisasmMemoryObject MemoryObject(Bytes, BytesSize, PC);
140
141  uint64_t Size;
142  MCInst Inst;
143  const MCDisassembler *DisAsm = DC->getDisAsm();
144  MCInstPrinter *IP = DC->getIP();
145  MCDisassembler::DecodeStatus S;
146  S = DisAsm->getInstruction(Inst, Size, MemoryObject, PC,
147                             /*REMOVE*/ nulls(), DC->CommentStream);
148  switch (S) {
149  case MCDisassembler::Fail:
150  case MCDisassembler::SoftFail:
151    // FIXME: Do something different for soft failure modes?
152    return 0;
153
154  case MCDisassembler::Success: {
155    DC->CommentStream.flush();
156    StringRef Comments = DC->CommentsToEmit.str();
157
158    SmallVector<char, 64> InsnStr;
159    raw_svector_ostream OS(InsnStr);
160    IP->printInst(&Inst, OS, Comments);
161    OS.flush();
162
163    // Tell the comment stream that the vector changed underneath it.
164    DC->CommentsToEmit.clear();
165    DC->CommentStream.resync();
166
167    assert(OutStringSize != 0 && "Output buffer cannot be zero size");
168    size_t OutputSize = std::min(OutStringSize-1, InsnStr.size());
169    std::memcpy(OutString, InsnStr.data(), OutputSize);
170    OutString[OutputSize] = '\0'; // Terminate string.
171
172    return Size;
173  }
174  }
175  return 0;
176}
177