1//===-- llvm-nm.cpp - Symbol table 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 traditional Unix "nm", that is, it
11// prints out the names of symbols in a bitcode or object file, along with some
12// information about each symbol.
13//
14// This "nm" supports many of the features of GNU "nm", including its different
15// output formats.
16//
17//===----------------------------------------------------------------------===//
18
19#include "llvm/LLVMContext.h"
20#include "llvm/Module.h"
21#include "llvm/Bitcode/ReaderWriter.h"
22#include "llvm/Bitcode/Archive.h"
23#include "llvm/Object/Archive.h"
24#include "llvm/Object/ObjectFile.h"
25#include "llvm/Support/CommandLine.h"
26#include "llvm/Support/FileSystem.h"
27#include "llvm/Support/ManagedStatic.h"
28#include "llvm/Support/MemoryBuffer.h"
29#include "llvm/Support/PrettyStackTrace.h"
30#include "llvm/Support/Program.h"
31#include "llvm/Support/raw_ostream.h"
32#include "llvm/Support/Signals.h"
33#include "llvm/Support/Format.h"
34#include "llvm/Support/system_error.h"
35#include <algorithm>
36#include <cctype>
37#include <cerrno>
38#include <cstring>
39#include <vector>
40using namespace llvm;
41using namespace object;
42
43namespace {
44  enum OutputFormatTy { bsd, sysv, posix };
45  cl::opt<OutputFormatTy>
46  OutputFormat("format",
47       cl::desc("Specify output format"),
48         cl::values(clEnumVal(bsd,   "BSD format"),
49                    clEnumVal(sysv,  "System V format"),
50                    clEnumVal(posix, "POSIX.2 format"),
51                    clEnumValEnd), cl::init(bsd));
52  cl::alias OutputFormat2("f", cl::desc("Alias for --format"),
53                          cl::aliasopt(OutputFormat));
54
55  cl::list<std::string>
56  InputFilenames(cl::Positional, cl::desc("<input bitcode files>"),
57                 cl::ZeroOrMore);
58
59  cl::opt<bool> UndefinedOnly("undefined-only",
60                              cl::desc("Show only undefined symbols"));
61  cl::alias UndefinedOnly2("u", cl::desc("Alias for --undefined-only"),
62                           cl::aliasopt(UndefinedOnly));
63
64  cl::opt<bool> DynamicSyms("dynamic",
65                             cl::desc("Display the dynamic symbols instead "
66                                      "of normal symbols."));
67  cl::alias DynamicSyms2("D", cl::desc("Alias for --dynamic"),
68                         cl::aliasopt(DynamicSyms));
69
70  cl::opt<bool> DefinedOnly("defined-only",
71                            cl::desc("Show only defined symbols"));
72
73  cl::opt<bool> ExternalOnly("extern-only",
74                             cl::desc("Show only external symbols"));
75  cl::alias ExternalOnly2("g", cl::desc("Alias for --extern-only"),
76                          cl::aliasopt(ExternalOnly));
77
78  cl::opt<bool> BSDFormat("B", cl::desc("Alias for --format=bsd"));
79  cl::opt<bool> POSIXFormat("P", cl::desc("Alias for --format=posix"));
80
81  cl::opt<bool> PrintFileName("print-file-name",
82    cl::desc("Precede each symbol with the object file it came from"));
83
84  cl::alias PrintFileNameA("A", cl::desc("Alias for --print-file-name"),
85                                cl::aliasopt(PrintFileName));
86  cl::alias PrintFileNameo("o", cl::desc("Alias for --print-file-name"),
87                                cl::aliasopt(PrintFileName));
88
89  cl::opt<bool> DebugSyms("debug-syms",
90    cl::desc("Show all symbols, even debugger only"));
91  cl::alias DebugSymsa("a", cl::desc("Alias for --debug-syms"),
92                            cl::aliasopt(DebugSyms));
93
94  cl::opt<bool> NumericSort("numeric-sort",
95    cl::desc("Sort symbols by address"));
96  cl::alias NumericSortn("n", cl::desc("Alias for --numeric-sort"),
97                              cl::aliasopt(NumericSort));
98  cl::alias NumericSortv("v", cl::desc("Alias for --numeric-sort"),
99                              cl::aliasopt(NumericSort));
100
101  cl::opt<bool> NoSort("no-sort",
102    cl::desc("Show symbols in order encountered"));
103  cl::alias NoSortp("p", cl::desc("Alias for --no-sort"),
104                         cl::aliasopt(NoSort));
105
106  cl::opt<bool> PrintSize("print-size",
107    cl::desc("Show symbol size instead of address"));
108  cl::alias PrintSizeS("S", cl::desc("Alias for --print-size"),
109                            cl::aliasopt(PrintSize));
110
111  cl::opt<bool> SizeSort("size-sort", cl::desc("Sort symbols by size"));
112
113  cl::opt<bool> WithoutAliases("without-aliases", cl::Hidden,
114                               cl::desc("Exclude aliases from output"));
115
116  bool PrintAddress = true;
117
118  bool MultipleFiles = false;
119
120  std::string ToolName;
121}
122
123
124static void error(Twine message, Twine path = Twine()) {
125  errs() << ToolName << ": " << path << ": " << message << ".\n";
126}
127
128static bool error(error_code ec, Twine path = Twine()) {
129  if (ec) {
130    error(ec.message(), path);
131    return true;
132  }
133  return false;
134}
135
136namespace {
137  struct NMSymbol {
138    uint64_t  Address;
139    uint64_t  Size;
140    char      TypeChar;
141    StringRef Name;
142  };
143
144  static bool CompareSymbolAddress(const NMSymbol &a, const NMSymbol &b) {
145    if (a.Address < b.Address)
146      return true;
147    else if (a.Address == b.Address && a.Name < b.Name)
148      return true;
149    else
150      return false;
151
152  }
153
154  static bool CompareSymbolSize(const NMSymbol &a, const NMSymbol &b) {
155    if (a.Size < b.Size)
156      return true;
157    else if (a.Size == b.Size && a.Name < b.Name)
158      return true;
159    else
160      return false;
161  }
162
163  static bool CompareSymbolName(const NMSymbol &a, const NMSymbol &b) {
164    return a.Name < b.Name;
165  }
166
167  StringRef CurrentFilename;
168  typedef std::vector<NMSymbol> SymbolListT;
169  SymbolListT SymbolList;
170}
171
172static void SortAndPrintSymbolList() {
173  if (!NoSort) {
174    if (NumericSort)
175      std::sort(SymbolList.begin(), SymbolList.end(), CompareSymbolAddress);
176    else if (SizeSort)
177      std::sort(SymbolList.begin(), SymbolList.end(), CompareSymbolSize);
178    else
179      std::sort(SymbolList.begin(), SymbolList.end(), CompareSymbolName);
180  }
181
182  if (OutputFormat == posix && MultipleFiles) {
183    outs() << '\n' << CurrentFilename << ":\n";
184  } else if (OutputFormat == bsd && MultipleFiles) {
185    outs() << "\n" << CurrentFilename << ":\n";
186  } else if (OutputFormat == sysv) {
187    outs() << "\n\nSymbols from " << CurrentFilename << ":\n\n"
188           << "Name                  Value   Class        Type"
189           << "         Size   Line  Section\n";
190  }
191
192  for (SymbolListT::iterator i = SymbolList.begin(),
193                             e = SymbolList.end(); i != e; ++i) {
194    if ((i->TypeChar != 'U') && UndefinedOnly)
195      continue;
196    if ((i->TypeChar == 'U') && DefinedOnly)
197      continue;
198    if (SizeSort && !PrintAddress && i->Size == UnknownAddressOrSize)
199      continue;
200
201    char SymbolAddrStr[10] = "";
202    char SymbolSizeStr[10] = "";
203
204    if (OutputFormat == sysv || i->Address == object::UnknownAddressOrSize)
205      strcpy(SymbolAddrStr, "        ");
206    if (OutputFormat == sysv)
207      strcpy(SymbolSizeStr, "        ");
208
209    if (i->Address != object::UnknownAddressOrSize)
210      format("%08" PRIx64, i->Address).print(SymbolAddrStr,
211                                             sizeof(SymbolAddrStr));
212    if (i->Size != object::UnknownAddressOrSize)
213      format("%08" PRIx64, i->Size).print(SymbolSizeStr, sizeof(SymbolSizeStr));
214
215    if (OutputFormat == posix) {
216      outs() << i->Name << " " << i->TypeChar << " "
217             << SymbolAddrStr << SymbolSizeStr << "\n";
218    } else if (OutputFormat == bsd) {
219      if (PrintAddress)
220        outs() << SymbolAddrStr << ' ';
221      if (PrintSize) {
222        outs() << SymbolSizeStr;
223        if (i->Size != object::UnknownAddressOrSize)
224          outs() << ' ';
225      }
226      outs() << i->TypeChar << " " << i->Name  << "\n";
227    } else if (OutputFormat == sysv) {
228      std::string PaddedName (i->Name);
229      while (PaddedName.length () < 20)
230        PaddedName += " ";
231      outs() << PaddedName << "|" << SymbolAddrStr << "|   "
232             << i->TypeChar
233             << "  |                  |" << SymbolSizeStr << "|     |\n";
234    }
235  }
236
237  SymbolList.clear();
238}
239
240static char TypeCharForSymbol(GlobalValue &GV) {
241  if (GV.isDeclaration())                                  return 'U';
242  if (GV.hasLinkOnceLinkage())                             return 'C';
243  if (GV.hasCommonLinkage())                               return 'C';
244  if (GV.hasWeakLinkage())                                 return 'W';
245  if (isa<Function>(GV) && GV.hasInternalLinkage())        return 't';
246  if (isa<Function>(GV))                                   return 'T';
247  if (isa<GlobalVariable>(GV) && GV.hasInternalLinkage())  return 'd';
248  if (isa<GlobalVariable>(GV))                             return 'D';
249  if (const GlobalAlias *GA = dyn_cast<GlobalAlias>(&GV)) {
250    const GlobalValue *AliasedGV = GA->getAliasedGlobal();
251    if (isa<Function>(AliasedGV))                          return 'T';
252    if (isa<GlobalVariable>(AliasedGV))                    return 'D';
253  }
254                                                           return '?';
255}
256
257static void DumpSymbolNameForGlobalValue(GlobalValue &GV) {
258  // Private linkage and available_externally linkage don't exist in symtab.
259  if (GV.hasPrivateLinkage() ||
260      GV.hasLinkerPrivateLinkage() ||
261      GV.hasLinkerPrivateWeakLinkage() ||
262      GV.hasAvailableExternallyLinkage())
263    return;
264  char TypeChar = TypeCharForSymbol(GV);
265  if (GV.hasLocalLinkage () && ExternalOnly)
266    return;
267
268  NMSymbol s;
269  s.Address = object::UnknownAddressOrSize;
270  s.Size = object::UnknownAddressOrSize;
271  s.TypeChar = TypeChar;
272  s.Name     = GV.getName();
273  SymbolList.push_back(s);
274}
275
276static void DumpSymbolNamesFromModule(Module *M) {
277  CurrentFilename = M->getModuleIdentifier();
278  std::for_each (M->begin(), M->end(), DumpSymbolNameForGlobalValue);
279  std::for_each (M->global_begin(), M->global_end(),
280                 DumpSymbolNameForGlobalValue);
281  if (!WithoutAliases)
282    std::for_each (M->alias_begin(), M->alias_end(),
283		   DumpSymbolNameForGlobalValue);
284
285  SortAndPrintSymbolList();
286}
287
288static void DumpSymbolNamesFromObject(ObjectFile *obj) {
289  error_code ec;
290  symbol_iterator ibegin = obj->begin_symbols();
291  symbol_iterator iend = obj->end_symbols();
292  if (DynamicSyms) {
293    ibegin = obj->begin_dynamic_symbols();
294    iend = obj->end_dynamic_symbols();
295  }
296  for (symbol_iterator i = ibegin; i != iend; i.increment(ec)) {
297    if (error(ec)) break;
298    uint32_t symflags;
299    if (error(i->getFlags(symflags))) break;
300    if (!DebugSyms && (symflags & SymbolRef::SF_FormatSpecific))
301      continue;
302    NMSymbol s;
303    s.Size = object::UnknownAddressOrSize;
304    s.Address = object::UnknownAddressOrSize;
305    if (PrintSize || SizeSort) {
306      if (error(i->getSize(s.Size))) break;
307    }
308    if (PrintAddress)
309      if (error(i->getAddress(s.Address))) break;
310    if (error(i->getNMTypeChar(s.TypeChar))) break;
311    if (error(i->getName(s.Name))) break;
312    SymbolList.push_back(s);
313  }
314
315  CurrentFilename = obj->getFileName();
316  SortAndPrintSymbolList();
317}
318
319static void DumpSymbolNamesFromFile(std::string &Filename) {
320  if (Filename != "-" && !sys::fs::exists(Filename)) {
321    errs() << ToolName << ": '" << Filename << "': " << "No such file\n";
322    return;
323  }
324
325  OwningPtr<MemoryBuffer> Buffer;
326  if (error(MemoryBuffer::getFileOrSTDIN(Filename, Buffer), Filename))
327    return;
328
329  sys::fs::file_magic magic = sys::fs::identify_magic(Buffer->getBuffer());
330
331  LLVMContext &Context = getGlobalContext();
332  std::string ErrorMessage;
333  if (magic == sys::fs::file_magic::bitcode) {
334    Module *Result = 0;
335    Result = ParseBitcodeFile(Buffer.get(), Context, &ErrorMessage);
336    if (Result) {
337      DumpSymbolNamesFromModule(Result);
338      delete Result;
339    } else {
340      error(ErrorMessage, Filename);
341      return;
342    }
343  } else if (magic == sys::fs::file_magic::archive) {
344    OwningPtr<Binary> arch;
345    if (error(object::createBinary(Buffer.take(), arch), Filename))
346      return;
347
348    if (object::Archive *a = dyn_cast<object::Archive>(arch.get())) {
349      for (object::Archive::child_iterator i = a->begin_children(),
350                                           e = a->end_children(); i != e; ++i) {
351        OwningPtr<Binary> child;
352        if (i->getAsBinary(child)) {
353          // Try opening it as a bitcode file.
354          OwningPtr<MemoryBuffer> buff(i->getBuffer());
355          Module *Result = 0;
356          if (buff)
357            Result = ParseBitcodeFile(buff.get(), Context, &ErrorMessage);
358
359          if (Result) {
360            DumpSymbolNamesFromModule(Result);
361            delete Result;
362          }
363          continue;
364        }
365        if (object::ObjectFile *o = dyn_cast<ObjectFile>(child.get())) {
366          outs() << o->getFileName() << ":\n";
367          DumpSymbolNamesFromObject(o);
368        }
369      }
370    }
371  } else if (magic.is_object()) {
372    OwningPtr<Binary> obj;
373    if (error(object::createBinary(Buffer.take(), obj), Filename))
374      return;
375    if (object::ObjectFile *o = dyn_cast<ObjectFile>(obj.get()))
376      DumpSymbolNamesFromObject(o);
377  } else {
378    errs() << ToolName << ": " << Filename << ": "
379           << "unrecognizable file type\n";
380    return;
381  }
382}
383
384int main(int argc, char **argv) {
385  // Print a stack trace if we signal out.
386  sys::PrintStackTraceOnErrorSignal();
387  PrettyStackTraceProgram X(argc, argv);
388
389  llvm_shutdown_obj Y;  // Call llvm_shutdown() on exit.
390  cl::ParseCommandLineOptions(argc, argv, "llvm symbol table dumper\n");
391
392  // llvm-nm only reads binary files.
393  if (error(sys::Program::ChangeStdinToBinary()))
394    return 1;
395
396  ToolName = argv[0];
397  if (BSDFormat) OutputFormat = bsd;
398  if (POSIXFormat) OutputFormat = posix;
399
400  // The relative order of these is important. If you pass --size-sort it should
401  // only print out the size. However, if you pass -S --size-sort, it should
402  // print out both the size and address.
403  if (SizeSort && !PrintSize) PrintAddress = false;
404  if (OutputFormat == sysv || SizeSort) PrintSize = true;
405
406  switch (InputFilenames.size()) {
407  case 0: InputFilenames.push_back("-");
408  case 1: break;
409  default: MultipleFiles = true;
410  }
411
412  std::for_each(InputFilenames.begin(), InputFilenames.end(),
413                DumpSymbolNamesFromFile);
414  return 0;
415}
416