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 #include "llvm/MC/MCAsmInfo.h"
13 #include "llvm/MC/MCContext.h"
14 #include "llvm/MC/MCDisassembler.h"
15 #include "llvm/MC/MCInst.h"
16 #include "llvm/MC/MCInstPrinter.h"
17 #include "llvm/MC/MCInstrInfo.h"
18 #include "llvm/MC/MCRegisterInfo.h"
19 #include "llvm/MC/MCSubtargetInfo.h"
20 #include "llvm/Support/ErrorHandling.h"
21 #include "llvm/Support/MemoryObject.h"
22 #include "llvm/Support/TargetRegistry.h"
23 
24 namespace llvm {
25 class Target;
26 } // namespace llvm
27 using 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 //
36 LLVMDisasmContextRef LLVMCreateDisasm(const char *TripleName, void *DisInfo,
37                                       int TagType, LLVMOpInfoCallback GetOpInfo,
38                                       LLVMSymbolLookupCallback SymbolLookUp) {
39   // Get the target.
40   std::string Error;
41   const Target *TheTarget = TargetRegistry::lookupTarget(TripleName, Error);
42   assert(TheTarget && "Unable to create target!");
43 
44   // Get the assembler info needed to setup the MCContext.
45   const MCAsmInfo *MAI = TheTarget->createMCAsmInfo(TripleName);
46   assert(MAI && "Unable to create target asm info!");
47 
48   const MCInstrInfo *MII = TheTarget->createMCInstrInfo();
49   assert(MII && "Unable to create target instruction info!");
50 
51   const MCRegisterInfo *MRI = TheTarget->createMCRegInfo(TripleName);
52   assert(MRI && "Unable to create target register info!");
53 
54   // Package up features to be passed to target/subtarget
55   std::string FeaturesStr;
56   std::string CPU;
57 
58   const MCSubtargetInfo *STI = TheTarget->createMCSubtargetInfo(TripleName, CPU,
59                                                                 FeaturesStr);
60   assert(STI && "Unable to create subtarget info!");
61 
62   // Set up the MCContext for creating symbols and MCExpr's.
63   MCContext *Ctx = new MCContext(*MAI, *MRI, 0);
64   assert(Ctx && "Unable to create MCContext!");
65 
66   // Set up disassembler.
67   MCDisassembler *DisAsm = TheTarget->createMCDisassembler(*STI);
68   assert(DisAsm && "Unable to create disassembler!");
69   DisAsm->setupForSymbolicDisassembly(GetOpInfo, SymbolLookUp, DisInfo, Ctx);
70 
71   // Set up the instruction printer.
72   int AsmPrinterVariant = MAI->getAssemblerDialect();
73   MCInstPrinter *IP = TheTarget->createMCInstPrinter(AsmPrinterVariant,
74                                                      *MAI, *MII, *MRI, *STI);
75   assert(IP && "Unable to create instruction printer!");
76 
77   LLVMDisasmContext *DC = new LLVMDisasmContext(TripleName, DisInfo, TagType,
78                                                 GetOpInfo, SymbolLookUp,
79                                                 TheTarget, MAI, MRI,
80                                                 STI, MII, Ctx, DisAsm, IP);
81   assert(DC && "Allocation failure!");
82 
83   return DC;
84 }
85 
86 //
87 // LLVMDisasmDispose() disposes of the disassembler specified by the context.
88 //
89 void LLVMDisasmDispose(LLVMDisasmContextRef DCR){
90   LLVMDisasmContext *DC = (LLVMDisasmContext *)DCR;
91   delete DC;
92 }
93 
94 namespace {
95 //
96 // The memory object created by LLVMDisasmInstruction().
97 //
98 class DisasmMemoryObject : public MemoryObject {
99   uint8_t *Bytes;
100   uint64_t Size;
101   uint64_t BasePC;
102 public:
103   DisasmMemoryObject(uint8_t *bytes, uint64_t size, uint64_t basePC) :
104                      Bytes(bytes), Size(size), BasePC(basePC) {}
105 
106   uint64_t getBase() const { return BasePC; }
107   uint64_t getExtent() const { return Size; }
108 
109   int readByte(uint64_t Addr, uint8_t *Byte) const {
110     if (Addr - BasePC >= Size)
111       return -1;
112     *Byte = Bytes[Addr - BasePC];
113     return 0;
114   }
115 };
116 } // end anonymous namespace
117 
118 //
119 // LLVMDisasmInstruction() disassembles a single instruction using the
120 // disassembler context specified in the parameter DC.  The bytes of the
121 // instruction are specified in the parameter Bytes, and contains at least
122 // BytesSize number of bytes.  The instruction is at the address specified by
123 // the PC parameter.  If a valid instruction can be disassembled its string is
124 // returned indirectly in OutString which whos size is specified in the
125 // parameter OutStringSize.  This function returns the number of bytes in the
126 // instruction or zero if there was no valid instruction.  If this function
127 // returns zero the caller will have to pick how many bytes they want to step
128 // over by printing a .byte, .long etc. to continue.
129 //
130 size_t LLVMDisasmInstruction(LLVMDisasmContextRef DCR, uint8_t *Bytes,
131                              uint64_t BytesSize, uint64_t PC, char *OutString,
132                              size_t OutStringSize){
133   LLVMDisasmContext *DC = (LLVMDisasmContext *)DCR;
134   // Wrap the pointer to the Bytes, BytesSize and PC in a MemoryObject.
135   DisasmMemoryObject MemoryObject(Bytes, BytesSize, PC);
136 
137   uint64_t Size;
138   MCInst Inst;
139   const MCDisassembler *DisAsm = DC->getDisAsm();
140   MCInstPrinter *IP = DC->getIP();
141   MCDisassembler::DecodeStatus S;
142   S = DisAsm->getInstruction(Inst, Size, MemoryObject, PC,
143                              /*REMOVE*/ nulls(), DC->CommentStream);
144   switch (S) {
145   case MCDisassembler::Fail:
146   case MCDisassembler::SoftFail:
147     // FIXME: Do something different for soft failure modes?
148     return 0;
149 
150   case MCDisassembler::Success: {
151     DC->CommentStream.flush();
152     StringRef Comments = DC->CommentsToEmit.str();
153 
154     SmallVector<char, 64> InsnStr;
155     raw_svector_ostream OS(InsnStr);
156     IP->printInst(&Inst, OS, Comments);
157     OS.flush();
158 
159     // Tell the comment stream that the vector changed underneath it.
160     DC->CommentsToEmit.clear();
161     DC->CommentStream.resync();
162 
163     assert(OutStringSize != 0 && "Output buffer cannot be zero size");
164     size_t OutputSize = std::min(OutStringSize-1, InsnStr.size());
165     std::memcpy(OutString, InsnStr.data(), OutputSize);
166     OutString[OutputSize] = '\0'; // Terminate string.
167 
168     return Size;
169   }
170   }
171   llvm_unreachable("Invalid DecodeStatus!");
172 }
173 
174 //
175 // LLVMSetDisasmOptions() sets the disassembler's options.  It returns 1 if it
176 // can set all the Options and 0 otherwise.
177 //
178 int LLVMSetDisasmOptions(LLVMDisasmContextRef DCR, uint64_t Options){
179   if (Options & LLVMDisassembler_Option_UseMarkup){
180       LLVMDisasmContext *DC = (LLVMDisasmContext *)DCR;
181       MCInstPrinter *IP = DC->getIP();
182       IP->setUseMarkup(1);
183       Options &= ~LLVMDisassembler_Option_UseMarkup;
184   }
185   if (Options & LLVMDisassembler_Option_PrintImmHex){
186       LLVMDisasmContext *DC = (LLVMDisasmContext *)DCR;
187       MCInstPrinter *IP = DC->getIP();
188       IP->setPrintImmHex(1);
189       Options &= ~LLVMDisassembler_Option_PrintImmHex;
190   }
191   return (Options == 0);
192 }
193