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 #include "Disassembler.h"
10 #include <stdio.h>
11 #include "llvm-c/Disassembler.h"
12 
13 #include <string>
14 #include "llvm/MC/MCAsmInfo.h"
15 #include "llvm/MC/MCDisassembler.h"
16 #include "llvm/MC/MCInst.h"
17 #include "llvm/MC/MCInstPrinter.h"
18 #include "llvm/MC/MCContext.h"
19 #include "llvm/Target/TargetRegistry.h"
20 #include "llvm/Target/TargetAsmInfo.h"  // FIXME.
21 #include "llvm/Target/TargetMachine.h"  // FIXME.
22 #include "llvm/Target/TargetSelect.h"
23 #include "llvm/Support/MemoryObject.h"
24 
25 namespace llvm {
26 class Target;
27 } // namespace llvm
28 using namespace llvm;
29 
30 #ifdef __cplusplus
31 extern "C" {
32 #endif // __cplusplus
33 
34 //
35 // LLVMCreateDisasm() creates a disassembler for the TripleName.  Symbolic
36 // disassembly is supported by passing a block of information in the DisInfo
37 // parameter and specifing the TagType and call back functions as described in
38 // the header llvm-c/Disassembler.h .  The pointer to the block and the
39 // functions can all be passed as NULL.  If successful this returns a
40 // disassembler context if not it returns NULL.
41 //
42 LLVMDisasmContextRef LLVMCreateDisasm(const char *TripleName, void *DisInfo,
43                                       int TagType, LLVMOpInfoCallback GetOpInfo,
44                                       LLVMSymbolLookupCallback SymbolLookUp) {
45   // Initialize targets and assembly printers/parsers.
46   llvm::InitializeAllTargetInfos();
47   // FIXME: We shouldn't need to initialize the Target(Machine)s.
48   llvm::InitializeAllTargets();
49   llvm::InitializeAllAsmPrinters();
50   llvm::InitializeAllAsmParsers();
51   llvm::InitializeAllDisassemblers();
52 
53   // Get the target.
54   std::string Error;
55   const Target *TheTarget = TargetRegistry::lookupTarget(TripleName, Error);
56   assert(TheTarget && "Unable to create target!");
57 
58   // Get the assembler info needed to setup the MCContext.
59   const MCAsmInfo *MAI = TheTarget->createAsmInfo(TripleName);
60   assert(MAI && "Unable to create target asm info!");
61 
62   // Package up features to be passed to target/subtarget
63   std::string FeaturesStr;
64 
65   // FIXME: We shouldn't need to do this (and link in codegen).
66   //        When we split this out, we should do it in a way that makes
67   //        it straightforward to switch subtargets on the fly.
68   TargetMachine *TM = TheTarget->createTargetMachine(TripleName, FeaturesStr);
69   assert(TM && "Unable to create target machine!");
70 
71   // Get the target assembler info needed to setup the context.
72   const TargetAsmInfo *tai = new TargetAsmInfo(*TM);
73   assert(tai && "Unable to create target assembler!");
74 
75   // Set up the MCContext for creating symbols and MCExpr's.
76   MCContext *Ctx = new MCContext(*MAI, tai);
77   assert(Ctx && "Unable to create MCContext!");
78 
79   // Set up disassembler.
80   MCDisassembler *DisAsm = TheTarget->createMCDisassembler();
81   assert(DisAsm && "Unable to create disassembler!");
82   DisAsm->setupForSymbolicDisassembly(GetOpInfo, DisInfo, Ctx);
83 
84   // Set up the instruction printer.
85   int AsmPrinterVariant = MAI->getAssemblerDialect();
86   MCInstPrinter *IP = TheTarget->createMCInstPrinter(*TM, AsmPrinterVariant,
87                                                      *MAI);
88   assert(IP && "Unable to create instruction printer!");
89 
90   LLVMDisasmContext *DC = new LLVMDisasmContext(TripleName, DisInfo, TagType,
91                                                 GetOpInfo, SymbolLookUp,
92                                                 TheTarget, MAI, TM, tai, Ctx,
93                                                 DisAsm, IP);
94   assert(DC && "Allocation failure!");
95   return DC;
96 }
97 
98 //
99 // LLVMDisasmDispose() disposes of the disassembler specified by the context.
100 //
101 void LLVMDisasmDispose(LLVMDisasmContextRef DCR){
102   LLVMDisasmContext *DC = (LLVMDisasmContext *)DCR;
103   delete DC;
104 }
105 
106 namespace {
107 //
108 // The memory object created by LLVMDisasmInstruction().
109 //
110 class DisasmMemoryObject : public MemoryObject {
111 private:
112   uint8_t *Bytes;
113   uint64_t Size;
114   uint64_t BasePC;
115 public:
116   DisasmMemoryObject(uint8_t *bytes, uint64_t size, uint64_t basePC) :
117                      Bytes(bytes), Size(size), BasePC(basePC) {}
118 
119   uint64_t getBase() const { return BasePC; }
120   uint64_t getExtent() const { return Size; }
121 
122   int readByte(uint64_t Addr, uint8_t *Byte) const {
123     if (Addr - BasePC >= Size)
124       return -1;
125     *Byte = Bytes[Addr - BasePC];
126     return 0;
127   }
128 };
129 } // namespace
130 
131 //
132 // LLVMDisasmInstruction() disassembles a single instruction using the
133 // disassembler context specified in the parameter DC.  The bytes of the
134 // instruction are specified in the parameter Bytes, and contains at least
135 // BytesSize number of bytes.  The instruction is at the address specified by
136 // the PC parameter.  If a valid instruction can be disassembled its string is
137 // returned indirectly in OutString which whos size is specified in the
138 // parameter OutStringSize.  This function returns the number of bytes in the
139 // instruction or zero if there was no valid instruction.  If this function
140 // returns zero the caller will have to pick how many bytes they want to step
141 // over by printing a .byte, .long etc. to continue.
142 //
143 size_t LLVMDisasmInstruction(LLVMDisasmContextRef DCR, uint8_t *Bytes,
144                              uint64_t BytesSize, uint64_t PC, char *OutString,
145                              size_t OutStringSize){
146   LLVMDisasmContext *DC = (LLVMDisasmContext *)DCR;
147   // Wrap the pointer to the Bytes, BytesSize and PC in a MemoryObject.
148   DisasmMemoryObject MemoryObject(Bytes, BytesSize, PC);
149 
150   uint64_t Size;
151   MCInst Inst;
152   const MCDisassembler *DisAsm = DC->getDisAsm();
153   MCInstPrinter *IP = DC->getIP();
154   if (!DisAsm->getInstruction(Inst, Size, MemoryObject, PC, /*REMOVE*/ nulls()))
155     return 0;
156 
157   std::string InsnStr;
158   raw_string_ostream OS(InsnStr);
159   IP->printInst(&Inst, OS);
160   OS.flush();
161 
162   size_t OutputSize = std::min(OutStringSize-1, InsnStr.size());
163   std::memcpy(OutString, InsnStr.data(), OutputSize);
164   OutString[OutputSize] = '\0'; // Terminate string.
165 
166   return Size;
167 }
168 
169 #ifdef __cplusplus
170 }
171 #endif // __cplusplus
172