Disassembler.cpp revision 276365dd4bc0c2160f91fd8062ae1fc90c86c324 
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
24namespace llvm {
25class Target;
26} // namespace llvm
27using 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//
36LLVMDisasmContextRef 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::InitializeAllAsmPrinters();
44  llvm::InitializeAllAsmParsers();
45  llvm::InitializeAllDisassemblers();
46
47  // Get the target.
48  std::string Error;
49  const Target *TheTarget = TargetRegistry::lookupTarget(TripleName, Error);
50  assert(TheTarget && "Unable to create target!");
51
52  // Get the assembler info needed to setup the MCContext.
53  const MCAsmInfo *MAI = TheTarget->createAsmInfo(TripleName);
54  assert(MAI && "Unable to create target asm info!");
55
56  // Package up features to be passed to target/subtarget
57  std::string FeaturesStr;
58  std::string CPU;
59
60  // FIXME: We shouldn't need to do this (and link in codegen).
61  //        When we split this out, we should do it in a way that makes
62  //        it straightforward to switch subtargets on the fly.
63  TargetMachine *TM = TheTarget->createTargetMachine(TripleName, CPU,
64                                                     FeaturesStr);
65  assert(TM && "Unable to create target machine!");
66
67  // Get the target assembler info needed to setup the context.
68  const TargetAsmInfo *tai = new TargetAsmInfo(*TM);
69  assert(tai && "Unable to create target assembler!");
70
71  // Set up the MCContext for creating symbols and MCExpr's.
72  MCContext *Ctx = new MCContext(*MAI, tai);
73  assert(Ctx && "Unable to create MCContext!");
74
75  // Set up disassembler.
76  MCDisassembler *DisAsm = TheTarget->createMCDisassembler();
77  assert(DisAsm && "Unable to create disassembler!");
78  DisAsm->setupForSymbolicDisassembly(GetOpInfo, DisInfo, Ctx);
79
80  // Set up the instruction printer.
81  int AsmPrinterVariant = MAI->getAssemblerDialect();
82  MCInstPrinter *IP = TheTarget->createMCInstPrinter(*TM, AsmPrinterVariant,
83                                                     *MAI);
84  assert(IP && "Unable to create instruction printer!");
85
86  LLVMDisasmContext *DC = new LLVMDisasmContext(TripleName, DisInfo, TagType,
87                                                GetOpInfo, SymbolLookUp,
88                                                TheTarget, MAI, TM, tai, Ctx,
89                                                DisAsm, IP);
90  assert(DC && "Allocation failure!");
91  return DC;
92}
93
94//
95// LLVMDisasmDispose() disposes of the disassembler specified by the context.
96//
97void LLVMDisasmDispose(LLVMDisasmContextRef DCR){
98  LLVMDisasmContext *DC = (LLVMDisasmContext *)DCR;
99  delete DC;
100}
101
102namespace {
103//
104// The memory object created by LLVMDisasmInstruction().
105//
106class DisasmMemoryObject : public MemoryObject {
107  uint8_t *Bytes;
108  uint64_t Size;
109  uint64_t BasePC;
110public:
111  DisasmMemoryObject(uint8_t *bytes, uint64_t size, uint64_t basePC) :
112                     Bytes(bytes), Size(size), BasePC(basePC) {}
113
114  uint64_t getBase() const { return BasePC; }
115  uint64_t getExtent() const { return Size; }
116
117  int readByte(uint64_t Addr, uint8_t *Byte) const {
118    if (Addr - BasePC >= Size)
119      return -1;
120    *Byte = Bytes[Addr - BasePC];
121    return 0;
122  }
123};
124} // end anonymous namespace
125
126//
127// LLVMDisasmInstruction() disassembles a single instruction using the
128// disassembler context specified in the parameter DC.  The bytes of the
129// instruction are specified in the parameter Bytes, and contains at least
130// BytesSize number of bytes.  The instruction is at the address specified by
131// the PC parameter.  If a valid instruction can be disassembled its string is
132// returned indirectly in OutString which whos size is specified in the
133// parameter OutStringSize.  This function returns the number of bytes in the
134// instruction or zero if there was no valid instruction.  If this function
135// returns zero the caller will have to pick how many bytes they want to step
136// over by printing a .byte, .long etc. to continue.
137//
138size_t LLVMDisasmInstruction(LLVMDisasmContextRef DCR, uint8_t *Bytes,
139                             uint64_t BytesSize, uint64_t PC, char *OutString,
140                             size_t OutStringSize){
141  LLVMDisasmContext *DC = (LLVMDisasmContext *)DCR;
142  // Wrap the pointer to the Bytes, BytesSize and PC in a MemoryObject.
143  DisasmMemoryObject MemoryObject(Bytes, BytesSize, PC);
144
145  uint64_t Size;
146  MCInst Inst;
147  const MCDisassembler *DisAsm = DC->getDisAsm();
148  MCInstPrinter *IP = DC->getIP();
149  if (!DisAsm->getInstruction(Inst, Size, MemoryObject, PC, /*REMOVE*/ nulls()))
150    return 0;
151
152  SmallVector<char, 64> InsnStr;
153  raw_svector_ostream OS(InsnStr);
154  IP->printInst(&Inst, OS);
155  OS.flush();
156
157  assert(OutStringSize != 0 && "Output buffer cannot be zero size");
158  size_t OutputSize = std::min(OutStringSize-1, InsnStr.size());
159  std::memcpy(OutString, InsnStr.data(), OutputSize);
160  OutString[OutputSize] = '\0'; // Terminate string.
161
162  return Size;
163}
164