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 // This config must be included before llvm-config.h.
18 #include "llvm/Config/config.h"
19 #include "../../lib/MC/MCDisassembler/EDDisassembler.h"
20 #include "../../lib/MC/MCDisassembler/EDInst.h"
21 #include "../../lib/MC/MCDisassembler/EDOperand.h"
22 #include "../../lib/MC/MCDisassembler/EDToken.h"
23 #include "llvm/ADT/OwningPtr.h"
24 #include "llvm/ADT/Triple.h"
25 #include "llvm/MC/MCAsmInfo.h"
26 #include "llvm/MC/MCDisassembler.h"
27 #include "llvm/MC/MCInst.h"
28 #include "llvm/MC/MCInstPrinter.h"
29 #include "llvm/Support/CommandLine.h"
30 #include "llvm/Support/Debug.h"
31 #include "llvm/Support/Format.h"
32 #include "llvm/Support/Host.h"
33 #include "llvm/Support/ManagedStatic.h"
34 #include "llvm/Support/MemoryBuffer.h"
35 #include "llvm/Support/MemoryObject.h"
36 #include "llvm/Support/PrettyStackTrace.h"
37 #include "llvm/Support/Signals.h"
38 #include "llvm/Support/SourceMgr.h"
39 #include "llvm/Support/raw_ostream.h"
40 #include "llvm/Support/system_error.h"
41 #include "llvm/Target/TargetMachine.h"
42 #include "llvm/Target/TargetRegistry.h"
43 #include "llvm/Target/TargetSelect.h"
44 #include <algorithm>
45 #include <cctype>
46 #include <cerrno>
47 #include <cstring>
48 using namespace llvm;
49 using namespace object;
50 
51 namespace {
52   cl::list<std::string>
53   InputFilenames(cl::Positional, cl::desc("<input object files>"),
54                  cl::ZeroOrMore);
55 
56   cl::opt<bool>
57   Disassemble("disassemble",
58     cl::desc("Display assembler mnemonics for the machine instructions"));
59   cl::alias
60   Disassembled("d", cl::desc("Alias for --disassemble"),
61                cl::aliasopt(Disassemble));
62 
63   cl::opt<std::string>
64   TripleName("triple", cl::desc("Target triple to disassemble for, "
65                                 "see -version for available targets"));
66 
67   cl::opt<std::string>
68   ArchName("arch", cl::desc("Target arch to disassemble for, "
69                             "see -version for available targets"));
70 
71   StringRef ToolName;
72 }
73 
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:
104   StringRefMemoryObject(StringRef bytes) : Bytes(bytes) {}
105 
106   uint64_t getBase() const { return 0; }
107   uint64_t getExtent() const { return Bytes.size(); }
108 
109   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 
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 
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\n";
163 
164   for (ObjectFile::section_iterator i = Obj->begin_sections(),
165                                     e = Obj->end_sections();
166                                     i != e; ++i) {
167     if (!i->isText())
168       continue;
169     outs() << "Disassembly of section " << i->getName() << ":\n\n";
170 
171     // Set up disassembler.
172     OwningPtr<const MCAsmInfo> AsmInfo(TheTarget->createAsmInfo(TripleName));
173 
174     if (!AsmInfo) {
175       errs() << "error: no assembly info for target " << TripleName << "\n";
176       return;
177     }
178 
179     OwningPtr<const MCDisassembler> DisAsm(TheTarget->createMCDisassembler());
180     if (!DisAsm) {
181       errs() << "error: no disassembler for target " << TripleName << "\n";
182       return;
183     }
184 
185     // FIXME: We shouldn't need to do this (and link in codegen).
186     //        When we split this out, we should do it in a way that makes
187     //        it straightforward to switch subtargets on the fly (.e.g,
188     //        the .cpu and .code16 directives).
189     std::string FeaturesStr;
190     OwningPtr<TargetMachine> TM(TheTarget->createTargetMachine(TripleName,
191                                                                FeaturesStr));
192     if (!TM) {
193       errs() << "error: could not create target for triple " << TripleName << "\n";
194       return;
195     }
196 
197     int AsmPrinterVariant = AsmInfo->getAssemblerDialect();
198     OwningPtr<MCInstPrinter> IP(TheTarget->createMCInstPrinter(
199                                   *TM, AsmPrinterVariant, *AsmInfo));
200     if (!IP) {
201       errs() << "error: no instruction printer for target " << TripleName << '\n';
202       return;
203     }
204 
205     StringRef Bytes = i->getContents();
206     StringRefMemoryObject memoryObject(Bytes);
207     uint64_t Size;
208     uint64_t Index;
209 
210     for (Index = 0; Index < Bytes.size(); Index += Size) {
211       MCInst Inst;
212 
213 #     ifndef NDEBUG
214       raw_ostream &DebugOut = DebugFlag ? dbgs() : nulls();
215 #     else
216       raw_ostream &DebugOut = nulls();
217 #     endif
218 
219       if (DisAsm->getInstruction(Inst, Size, memoryObject, Index, DebugOut)) {
220         outs() << format("%8x:\t", i->getAddress() + Index);
221         DumpBytes(StringRef(Bytes.data() + Index, Size));
222         IP->printInst(&Inst, outs());
223         outs() << "\n";
224       } else {
225         errs() << ToolName << ": warning: invalid instruction encoding\n";
226         if (Size == 0)
227           Size = 1; // skip illegible bytes
228       }
229     }
230   }
231 }
232 
233 int main(int argc, char **argv) {
234   // Print a stack trace if we signal out.
235   sys::PrintStackTraceOnErrorSignal();
236   PrettyStackTraceProgram X(argc, argv);
237   llvm_shutdown_obj Y;  // Call llvm_shutdown() on exit.
238 
239   // Initialize targets and assembly printers/parsers.
240   llvm::InitializeAllTargetInfos();
241   // FIXME: We shouldn't need to initialize the Target(Machine)s.
242   llvm::InitializeAllTargets();
243   llvm::InitializeAllAsmPrinters();
244   llvm::InitializeAllAsmParsers();
245   llvm::InitializeAllDisassemblers();
246 
247   cl::ParseCommandLineOptions(argc, argv, "llvm object file dumper\n");
248   TripleName = Triple::normalize(TripleName);
249 
250   ToolName = argv[0];
251 
252   // Defaults to a.out if no filenames specified.
253   if (InputFilenames.size() == 0)
254     InputFilenames.push_back("a.out");
255 
256   // -d is the only flag that is currently implemented, so just print help if
257   // it is not set.
258   if (!Disassemble) {
259     cl::PrintHelpMessage();
260     return 2;
261   }
262 
263   std::for_each(InputFilenames.begin(), InputFilenames.end(),
264                 DisassembleInput);
265 
266   return 0;
267 }
268