1 //===-- llvm-objdump.cpp - Object file dumping utility for llvm -----------===//
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 program is a utility that works like binutils "objdump", that is, it
11 // dumps out a plethora of information about an object file depending on the
12 // flags.
13 //
14 //===----------------------------------------------------------------------===//
15
16 #include "llvm/Object/ObjectFile.h"
17 #include "llvm/ADT/OwningPtr.h"
18 #include "llvm/ADT/Triple.h"
19 #include "llvm/ADT/STLExtras.h"
20 #include "llvm/MC/MCAsmInfo.h"
21 #include "llvm/MC/MCDisassembler.h"
22 #include "llvm/MC/MCInst.h"
23 #include "llvm/MC/MCInstPrinter.h"
24 #include "llvm/Support/CommandLine.h"
25 #include "llvm/Support/Debug.h"
26 #include "llvm/Support/Format.h"
27 #include "llvm/Support/Host.h"
28 #include "llvm/Support/ManagedStatic.h"
29 #include "llvm/Support/MemoryBuffer.h"
30 #include "llvm/Support/MemoryObject.h"
31 #include "llvm/Support/PrettyStackTrace.h"
32 #include "llvm/Support/Signals.h"
33 #include "llvm/Support/SourceMgr.h"
34 #include "llvm/Support/raw_ostream.h"
35 #include "llvm/Support/system_error.h"
36 #include "llvm/Target/TargetRegistry.h"
37 #include "llvm/Target/TargetSelect.h"
38 #include <algorithm>
39 #include <cstring>
40 using namespace llvm;
41 using namespace object;
42
43 namespace {
44 cl::list<std::string>
45 InputFilenames(cl::Positional, cl::desc("<input object files>"),
46 cl::ZeroOrMore);
47
48 cl::opt<bool>
49 Disassemble("disassemble",
50 cl::desc("Display assembler mnemonics for the machine instructions"));
51 cl::alias
52 Disassembled("d", cl::desc("Alias for --disassemble"),
53 cl::aliasopt(Disassemble));
54
55 cl::opt<std::string>
56 TripleName("triple", cl::desc("Target triple to disassemble for, "
57 "see -version for available targets"));
58
59 cl::opt<std::string>
60 ArchName("arch", cl::desc("Target arch to disassemble for, "
61 "see -version for available targets"));
62
63 StringRef ToolName;
64
error(error_code ec)65 bool error(error_code ec) {
66 if (!ec) return false;
67
68 outs() << ToolName << ": error reading file: " << ec.message() << ".\n";
69 outs().flush();
70 return true;
71 }
72 }
73
GetTarget(const ObjectFile * Obj=NULL)74 static const Target *GetTarget(const ObjectFile *Obj = NULL) {
75 // Figure out the target triple.
76 llvm::Triple TT("unknown-unknown-unknown");
77 if (TripleName.empty()) {
78 if (Obj)
79 TT.setArch(Triple::ArchType(Obj->getArch()));
80 } else
81 TT.setTriple(Triple::normalize(TripleName));
82
83 if (!ArchName.empty())
84 TT.setArchName(ArchName);
85
86 TripleName = TT.str();
87
88 // Get the target specific parser.
89 std::string Error;
90 const Target *TheTarget = TargetRegistry::lookupTarget(TripleName, Error);
91 if (TheTarget)
92 return TheTarget;
93
94 errs() << ToolName << ": error: unable to get target for '" << TripleName
95 << "', see --version and --triple.\n";
96 return 0;
97 }
98
99 namespace {
100 class StringRefMemoryObject : public MemoryObject {
101 private:
102 StringRef Bytes;
103 public:
StringRefMemoryObject(StringRef bytes)104 StringRefMemoryObject(StringRef bytes) : Bytes(bytes) {}
105
getBase() const106 uint64_t getBase() const { return 0; }
getExtent() const107 uint64_t getExtent() const { return Bytes.size(); }
108
readByte(uint64_t Addr,uint8_t * Byte) const109 int readByte(uint64_t Addr, uint8_t *Byte) const {
110 if (Addr >= getExtent())
111 return -1;
112 *Byte = Bytes[Addr];
113 return 0;
114 }
115 };
116 }
117
DumpBytes(StringRef bytes)118 static void DumpBytes(StringRef bytes) {
119 static char hex_rep[] = "0123456789abcdef";
120 // FIXME: The real way to do this is to figure out the longest instruction
121 // and align to that size before printing. I'll fix this when I get
122 // around to outputting relocations.
123 // 15 is the longest x86 instruction
124 // 3 is for the hex rep of a byte + a space.
125 // 1 is for the null terminator.
126 enum { OutputSize = (15 * 3) + 1 };
127 char output[OutputSize];
128
129 assert(bytes.size() <= 15
130 && "DumpBytes only supports instructions of up to 15 bytes");
131 memset(output, ' ', sizeof(output));
132 unsigned index = 0;
133 for (StringRef::iterator i = bytes.begin(),
134 e = bytes.end(); i != e; ++i) {
135 output[index] = hex_rep[(*i & 0xF0) >> 4];
136 output[index + 1] = hex_rep[*i & 0xF];
137 index += 3;
138 }
139
140 output[sizeof(output) - 1] = 0;
141 outs() << output;
142 }
143
DisassembleInput(const StringRef & Filename)144 static void DisassembleInput(const StringRef &Filename) {
145 OwningPtr<MemoryBuffer> Buff;
146
147 if (error_code ec = MemoryBuffer::getFileOrSTDIN(Filename, Buff)) {
148 errs() << ToolName << ": " << Filename << ": " << ec.message() << "\n";
149 return;
150 }
151
152 OwningPtr<ObjectFile> Obj(ObjectFile::createObjectFile(Buff.take()));
153
154 const Target *TheTarget = GetTarget(Obj.get());
155 if (!TheTarget) {
156 // GetTarget prints out stuff.
157 return;
158 }
159
160 outs() << '\n';
161 outs() << Filename
162 << ":\tfile format " << Obj->getFileFormatName() << "\n\n";
163
164 error_code ec;
165 for (ObjectFile::section_iterator i = Obj->begin_sections(),
166 e = Obj->end_sections();
167 i != e; i.increment(ec)) {
168 if (error(ec)) break;
169 bool text;
170 if (error(i->isText(text))) break;
171 if (!text) continue;
172
173 // Make a list of all the symbols in this section.
174 std::vector<std::pair<uint64_t, StringRef> > Symbols;
175 for (ObjectFile::symbol_iterator si = Obj->begin_symbols(),
176 se = Obj->end_symbols();
177 si != se; si.increment(ec)) {
178 bool contains;
179 if (!error(i->containsSymbol(*si, contains)) && contains) {
180 uint64_t Address;
181 if (error(si->getAddress(Address))) break;
182 StringRef Name;
183 if (error(si->getName(Name))) break;
184 Symbols.push_back(std::make_pair(Address, Name));
185 }
186 }
187
188 // Sort the symbols by address, just in case they didn't come in that way.
189 array_pod_sort(Symbols.begin(), Symbols.end());
190
191 StringRef name;
192 if (error(i->getName(name))) break;
193 outs() << "Disassembly of section " << name << ':';
194
195 // If the section has no symbols just insert a dummy one and disassemble
196 // the whole section.
197 if (Symbols.empty())
198 Symbols.push_back(std::make_pair(0, name));
199
200 // Set up disassembler.
201 OwningPtr<const MCAsmInfo> AsmInfo(TheTarget->createMCAsmInfo(TripleName));
202
203 if (!AsmInfo) {
204 errs() << "error: no assembly info for target " << TripleName << "\n";
205 return;
206 }
207
208 OwningPtr<const MCDisassembler> DisAsm(TheTarget->createMCDisassembler());
209 if (!DisAsm) {
210 errs() << "error: no disassembler for target " << TripleName << "\n";
211 return;
212 }
213
214 int AsmPrinterVariant = AsmInfo->getAssemblerDialect();
215 OwningPtr<MCInstPrinter> IP(TheTarget->createMCInstPrinter(
216 AsmPrinterVariant, *AsmInfo));
217 if (!IP) {
218 errs() << "error: no instruction printer for target " << TripleName << '\n';
219 return;
220 }
221
222 StringRef Bytes;
223 if (error(i->getContents(Bytes))) break;
224 StringRefMemoryObject memoryObject(Bytes);
225 uint64_t Size;
226 uint64_t Index;
227 uint64_t SectSize;
228 if (error(i->getSize(SectSize))) break;
229
230 // Disassemble symbol by symbol.
231 for (unsigned si = 0, se = Symbols.size(); si != se; ++si) {
232 uint64_t Start = Symbols[si].first;
233 uint64_t End = si == se-1 ? SectSize : Symbols[si + 1].first - 1;
234 outs() << '\n' << Symbols[si].second << ":\n";
235
236 for (Index = Start; Index < End; Index += Size) {
237 MCInst Inst;
238
239 #ifndef NDEBUG
240 raw_ostream &DebugOut = DebugFlag ? dbgs() : nulls();
241 #else
242 raw_ostream &DebugOut = nulls();
243 #endif
244
245 if (DisAsm->getInstruction(Inst, Size, memoryObject, Index, DebugOut)) {
246 uint64_t addr;
247 if (error(i->getAddress(addr))) break;
248 outs() << format("%8x:\t", addr + Index);
249 DumpBytes(StringRef(Bytes.data() + Index, Size));
250 IP->printInst(&Inst, outs());
251 outs() << "\n";
252 } else {
253 errs() << ToolName << ": warning: invalid instruction encoding\n";
254 if (Size == 0)
255 Size = 1; // skip illegible bytes
256 }
257 }
258 }
259 }
260 }
261
main(int argc,char ** argv)262 int main(int argc, char **argv) {
263 // Print a stack trace if we signal out.
264 sys::PrintStackTraceOnErrorSignal();
265 PrettyStackTraceProgram X(argc, argv);
266 llvm_shutdown_obj Y; // Call llvm_shutdown() on exit.
267
268 // Initialize targets and assembly printers/parsers.
269 llvm::InitializeAllTargetInfos();
270 // FIXME: We shouldn't need to initialize the Target(Machine)s.
271 llvm::InitializeAllTargets();
272 llvm::InitializeAllMCAsmInfos();
273 llvm::InitializeAllMCCodeGenInfos();
274 llvm::InitializeAllAsmPrinters();
275 llvm::InitializeAllAsmParsers();
276 llvm::InitializeAllDisassemblers();
277
278 cl::ParseCommandLineOptions(argc, argv, "llvm object file dumper\n");
279 TripleName = Triple::normalize(TripleName);
280
281 ToolName = argv[0];
282
283 // Defaults to a.out if no filenames specified.
284 if (InputFilenames.size() == 0)
285 InputFilenames.push_back("a.out");
286
287 // -d is the only flag that is currently implemented, so just print help if
288 // it is not set.
289 if (!Disassemble) {
290 cl::PrintHelpMessage();
291 return 2;
292 }
293
294 std::for_each(InputFilenames.begin(), InputFilenames.end(),
295 DisassembleInput);
296
297 return 0;
298 }
299