1 //===- Disassembler.cpp - Disassembler for hex strings --------------------===//
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 // This class implements the disassembler of strings of bytes written in
11 // hexadecimal, from standard input or from a file.
12 //
13 //===----------------------------------------------------------------------===//
14
15 #include "Disassembler.h"
16 #include "../../lib/MC/MCDisassembler/EDDisassembler.h"
17 #include "../../lib/MC/MCDisassembler/EDInst.h"
18 #include "../../lib/MC/MCDisassembler/EDOperand.h"
19 #include "../../lib/MC/MCDisassembler/EDToken.h"
20 #include "llvm/MC/MCDisassembler.h"
21 #include "llvm/MC/MCInst.h"
22 #include "llvm/MC/MCStreamer.h"
23 #include "llvm/MC/MCSubtargetInfo.h"
24 #include "llvm/ADT/OwningPtr.h"
25 #include "llvm/ADT/Triple.h"
26 #include "llvm/Support/MemoryBuffer.h"
27 #include "llvm/Support/MemoryObject.h"
28 #include "llvm/Support/SourceMgr.h"
29 #include "llvm/Support/TargetRegistry.h"
30 #include "llvm/Support/raw_ostream.h"
31
32 using namespace llvm;
33
34 typedef std::vector<std::pair<unsigned char, const char*> > ByteArrayTy;
35
36 namespace {
37 class VectorMemoryObject : public MemoryObject {
38 private:
39 const ByteArrayTy &Bytes;
40 public:
VectorMemoryObject(const ByteArrayTy & bytes)41 VectorMemoryObject(const ByteArrayTy &bytes) : Bytes(bytes) {}
42
getBase() const43 uint64_t getBase() const { return 0; }
getExtent() const44 uint64_t getExtent() const { return Bytes.size(); }
45
readByte(uint64_t Addr,uint8_t * Byte) const46 int readByte(uint64_t Addr, uint8_t *Byte) const {
47 if (Addr >= getExtent())
48 return -1;
49 *Byte = Bytes[Addr].first;
50 return 0;
51 }
52 };
53 }
54
PrintInsts(const MCDisassembler & DisAsm,const ByteArrayTy & Bytes,SourceMgr & SM,raw_ostream & Out,MCStreamer & Streamer)55 static bool PrintInsts(const MCDisassembler &DisAsm,
56 const ByteArrayTy &Bytes,
57 SourceMgr &SM, raw_ostream &Out,
58 MCStreamer &Streamer) {
59 // Wrap the vector in a MemoryObject.
60 VectorMemoryObject memoryObject(Bytes);
61
62 // Disassemble it to strings.
63 uint64_t Size;
64 uint64_t Index;
65
66 for (Index = 0; Index < Bytes.size(); Index += Size) {
67 MCInst Inst;
68
69 MCDisassembler::DecodeStatus S;
70 S = DisAsm.getInstruction(Inst, Size, memoryObject, Index,
71 /*REMOVE*/ nulls(), nulls());
72 switch (S) {
73 case MCDisassembler::Fail:
74 SM.PrintMessage(SMLoc::getFromPointer(Bytes[Index].second),
75 SourceMgr::DK_Warning,
76 "invalid instruction encoding");
77 if (Size == 0)
78 Size = 1; // skip illegible bytes
79 break;
80
81 case MCDisassembler::SoftFail:
82 SM.PrintMessage(SMLoc::getFromPointer(Bytes[Index].second),
83 SourceMgr::DK_Warning,
84 "potentially undefined instruction encoding");
85 // Fall through
86
87 case MCDisassembler::Success:
88 Streamer.EmitInstruction(Inst);
89 break;
90 }
91 }
92
93 return false;
94 }
95
ByteArrayFromString(ByteArrayTy & ByteArray,StringRef & Str,SourceMgr & SM)96 static bool ByteArrayFromString(ByteArrayTy &ByteArray,
97 StringRef &Str,
98 SourceMgr &SM) {
99 while (!Str.empty()) {
100 // Strip horizontal whitespace.
101 if (size_t Pos = Str.find_first_not_of(" \t\r")) {
102 Str = Str.substr(Pos);
103 continue;
104 }
105
106 // If this is the end of a line or start of a comment, remove the rest of
107 // the line.
108 if (Str[0] == '\n' || Str[0] == '#') {
109 // Strip to the end of line if we already processed any bytes on this
110 // line. This strips the comment and/or the \n.
111 if (Str[0] == '\n') {
112 Str = Str.substr(1);
113 } else {
114 Str = Str.substr(Str.find_first_of('\n'));
115 if (!Str.empty())
116 Str = Str.substr(1);
117 }
118 continue;
119 }
120
121 // Get the current token.
122 size_t Next = Str.find_first_of(" \t\n\r#");
123 StringRef Value = Str.substr(0, Next);
124
125 // Convert to a byte and add to the byte vector.
126 unsigned ByteVal;
127 if (Value.getAsInteger(0, ByteVal) || ByteVal > 255) {
128 // If we have an error, print it and skip to the end of line.
129 SM.PrintMessage(SMLoc::getFromPointer(Value.data()), SourceMgr::DK_Error,
130 "invalid input token");
131 Str = Str.substr(Str.find('\n'));
132 ByteArray.clear();
133 continue;
134 }
135
136 ByteArray.push_back(std::make_pair((unsigned char)ByteVal, Value.data()));
137 Str = Str.substr(Next);
138 }
139
140 return false;
141 }
142
disassemble(const Target & T,const std::string & Triple,MCSubtargetInfo & STI,MCStreamer & Streamer,MemoryBuffer & Buffer,SourceMgr & SM,raw_ostream & Out)143 int Disassembler::disassemble(const Target &T,
144 const std::string &Triple,
145 MCSubtargetInfo &STI,
146 MCStreamer &Streamer,
147 MemoryBuffer &Buffer,
148 SourceMgr &SM,
149 raw_ostream &Out) {
150 OwningPtr<const MCDisassembler> DisAsm(T.createMCDisassembler(STI));
151 if (!DisAsm) {
152 errs() << "error: no disassembler for target " << Triple << "\n";
153 return -1;
154 }
155
156 // Set up initial section manually here
157 Streamer.InitSections();
158
159 bool ErrorOccurred = false;
160
161 // Convert the input to a vector for disassembly.
162 ByteArrayTy ByteArray;
163 StringRef Str = Buffer.getBuffer();
164
165 ErrorOccurred |= ByteArrayFromString(ByteArray, Str, SM);
166
167 if (!ByteArray.empty())
168 ErrorOccurred |= PrintInsts(*DisAsm, ByteArray, SM, Out, Streamer);
169
170 return ErrorOccurred;
171 }
172
byteArrayReader(uint8_t * B,uint64_t A,void * Arg)173 static int byteArrayReader(uint8_t *B, uint64_t A, void *Arg) {
174 ByteArrayTy &ByteArray = *((ByteArrayTy*)Arg);
175
176 if (A >= ByteArray.size())
177 return -1;
178
179 *B = ByteArray[A].first;
180
181 return 0;
182 }
183
verboseEvaluator(uint64_t * V,unsigned R,void * Arg)184 static int verboseEvaluator(uint64_t *V, unsigned R, void *Arg) {
185 EDDisassembler &disassembler = *(EDDisassembler *)((void **)Arg)[0];
186 raw_ostream &Out = *(raw_ostream *)((void **)Arg)[1];
187
188 if (const char *regName = disassembler.nameWithRegisterID(R))
189 Out << "[" << regName << "/" << R << "]";
190
191 if (disassembler.registerIsStackPointer(R))
192 Out << "(sp)";
193 if (disassembler.registerIsProgramCounter(R))
194 Out << "(pc)";
195
196 *V = 0;
197 return 0;
198 }
199
disassembleEnhanced(const std::string & TS,MemoryBuffer & Buffer,SourceMgr & SM,raw_ostream & Out)200 int Disassembler::disassembleEnhanced(const std::string &TS,
201 MemoryBuffer &Buffer,
202 SourceMgr &SM,
203 raw_ostream &Out) {
204 ByteArrayTy ByteArray;
205 StringRef Str = Buffer.getBuffer();
206
207 if (ByteArrayFromString(ByteArray, Str, SM)) {
208 return -1;
209 }
210
211 Triple T(TS);
212 EDDisassembler::AssemblySyntax AS;
213
214 switch (T.getArch()) {
215 default:
216 errs() << "error: no default assembly syntax for " << TS.c_str() << "\n";
217 return -1;
218 case Triple::arm:
219 case Triple::thumb:
220 AS = EDDisassembler::kEDAssemblySyntaxARMUAL;
221 break;
222 case Triple::x86:
223 case Triple::x86_64:
224 AS = EDDisassembler::kEDAssemblySyntaxX86ATT;
225 break;
226 }
227
228 OwningPtr<EDDisassembler>
229 disassembler(EDDisassembler::getDisassembler(TS.c_str(), AS));
230
231 if (disassembler == 0) {
232 errs() << "error: couldn't get disassembler for " << TS << '\n';
233 return -1;
234 }
235
236 while (ByteArray.size()) {
237 OwningPtr<EDInst>
238 inst(disassembler->createInst(byteArrayReader, 0, &ByteArray));
239
240 if (inst == 0) {
241 errs() << "error: Didn't get an instruction\n";
242 return -1;
243 }
244
245 ByteArray.erase (ByteArray.begin(), ByteArray.begin() + inst->byteSize());
246
247 unsigned numTokens = inst->numTokens();
248 if ((int)numTokens < 0) {
249 errs() << "error: couldn't count the instruction's tokens\n";
250 return -1;
251 }
252
253 for (unsigned tokenIndex = 0; tokenIndex != numTokens; ++tokenIndex) {
254 EDToken *token;
255
256 if (inst->getToken(token, tokenIndex)) {
257 errs() << "error: Couldn't get token\n";
258 return -1;
259 }
260
261 const char *buf;
262 if (token->getString(buf)) {
263 errs() << "error: Couldn't get string for token\n";
264 return -1;
265 }
266
267 Out << '[';
268 int operandIndex = token->operandID();
269
270 if (operandIndex >= 0)
271 Out << operandIndex << "-";
272
273 switch (token->type()) {
274 case EDToken::kTokenWhitespace: Out << "w"; break;
275 case EDToken::kTokenPunctuation: Out << "p"; break;
276 case EDToken::kTokenOpcode: Out << "o"; break;
277 case EDToken::kTokenLiteral: Out << "l"; break;
278 case EDToken::kTokenRegister: Out << "r"; break;
279 }
280
281 Out << ":" << buf;
282
283 if (token->type() == EDToken::kTokenLiteral) {
284 Out << "=";
285 if (token->literalSign())
286 Out << "-";
287 uint64_t absoluteValue;
288 if (token->literalAbsoluteValue(absoluteValue)) {
289 errs() << "error: Couldn't get the value of a literal token\n";
290 return -1;
291 }
292 Out << absoluteValue;
293 } else if (token->type() == EDToken::kTokenRegister) {
294 Out << "=";
295 unsigned regID;
296 if (token->registerID(regID)) {
297 errs() << "error: Couldn't get the ID of a register token\n";
298 return -1;
299 }
300 Out << "r" << regID;
301 }
302
303 Out << "]";
304 }
305
306 Out << " ";
307
308 if (inst->isBranch())
309 Out << "<br> ";
310 if (inst->isMove())
311 Out << "<mov> ";
312
313 unsigned numOperands = inst->numOperands();
314
315 if ((int)numOperands < 0) {
316 errs() << "error: Couldn't count operands\n";
317 return -1;
318 }
319
320 for (unsigned operandIndex = 0; operandIndex != numOperands;
321 ++operandIndex) {
322 Out << operandIndex << ":";
323
324 EDOperand *operand;
325 if (inst->getOperand(operand, operandIndex)) {
326 errs() << "error: couldn't get operand\n";
327 return -1;
328 }
329
330 uint64_t evaluatedResult;
331 void *Arg[] = { disassembler.get(), &Out };
332 if (operand->evaluate(evaluatedResult, verboseEvaluator, Arg)) {
333 errs() << "error: Couldn't evaluate an operand\n";
334 return -1;
335 }
336 Out << "=" << evaluatedResult << " ";
337 }
338
339 Out << '\n';
340 }
341
342 return 0;
343 }
344