1 //===-- llvm-objdump.cpp - Object file dumping utility for llvm -----------===//
3 // The LLVM Compiler Infrastructure
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
8 //===----------------------------------------------------------------------===//
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
14 // The flags and output of this program should be near identical to those of
17 //===----------------------------------------------------------------------===//
19 #include "llvm-objdump.h"
20 #include "MCFunction.h"
21 #include "llvm/ADT/OwningPtr.h"
22 #include "llvm/ADT/STLExtras.h"
23 #include "llvm/ADT/StringExtras.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/MC/MCInstrInfo.h"
30 #include "llvm/MC/MCRegisterInfo.h"
31 #include "llvm/MC/MCSubtargetInfo.h"
32 #include "llvm/Object/Archive.h"
33 #include "llvm/Object/COFF.h"
34 #include "llvm/Object/MachO.h"
35 #include "llvm/Object/ObjectFile.h"
36 #include "llvm/Support/Casting.h"
37 #include "llvm/Support/CommandLine.h"
38 #include "llvm/Support/Debug.h"
39 #include "llvm/Support/FileSystem.h"
40 #include "llvm/Support/Format.h"
41 #include "llvm/Support/GraphWriter.h"
42 #include "llvm/Support/Host.h"
43 #include "llvm/Support/ManagedStatic.h"
44 #include "llvm/Support/MemoryBuffer.h"
45 #include "llvm/Support/MemoryObject.h"
46 #include "llvm/Support/PrettyStackTrace.h"
47 #include "llvm/Support/Signals.h"
48 #include "llvm/Support/SourceMgr.h"
49 #include "llvm/Support/TargetRegistry.h"
50 #include "llvm/Support/TargetSelect.h"
51 #include "llvm/Support/raw_ostream.h"
52 #include "llvm/Support/system_error.h"
57 using namespace object
;
59 static cl::list
<std::string
>
60 InputFilenames(cl::Positional
, cl::desc("<input object files>"),cl::ZeroOrMore
);
63 Disassemble("disassemble",
64 cl::desc("Display assembler mnemonics for the machine instructions"));
66 Disassembled("d", cl::desc("Alias for --disassemble"),
67 cl::aliasopt(Disassemble
));
70 Relocations("r", cl::desc("Display the relocation entries in the file"));
73 SectionContents("s", cl::desc("Display the content of each section"));
76 SymbolTable("t", cl::desc("Display the symbol table"));
79 MachOOpt("macho", cl::desc("Use MachO specific object file parser"));
81 MachOm("m", cl::desc("Alias for --macho"), cl::aliasopt(MachOOpt
));
84 llvm::TripleName("triple", cl::desc("Target triple to disassemble for, "
85 "see -version for available targets"));
88 llvm::ArchName("arch", cl::desc("Target arch to disassemble for, "
89 "see -version for available targets"));
92 SectionHeaders("section-headers", cl::desc("Display summaries of the headers "
93 "for each section."));
95 SectionHeadersShort("headers", cl::desc("Alias for --section-headers"),
96 cl::aliasopt(SectionHeaders
));
98 SectionHeadersShorter("h", cl::desc("Alias for --section-headers"),
99 cl::aliasopt(SectionHeaders
));
101 static cl::list
<std::string
>
104 cl::desc("Target specific attributes"),
105 cl::value_desc("a1,+a2,-a3,..."));
108 NoShowRawInsn("no-show-raw-insn", cl::desc("When disassembling instructions, "
109 "do not print the instruction bytes."));
112 UnwindInfo("unwind-info", cl::desc("Display unwind information"));
115 UnwindInfoShort("u", cl::desc("Alias for --unwind-info"),
116 cl::aliasopt(UnwindInfo
));
119 PrivateHeaders("private-headers",
120 cl::desc("Display format specific file headers"));
123 PrivateHeadersShort("p", cl::desc("Alias for --private-headers"),
124 cl::aliasopt(PrivateHeaders
));
126 static StringRef ToolName
;
128 bool llvm::error(error_code ec
) {
129 if (!ec
) return false;
131 outs() << ToolName
<< ": error reading file: " << ec
.message() << ".\n";
136 static const Target
*getTarget(const ObjectFile
*Obj
= NULL
) {
137 // Figure out the target triple.
138 llvm::Triple
TheTriple("unknown-unknown-unknown");
139 if (TripleName
.empty()) {
141 TheTriple
.setArch(Triple::ArchType(Obj
->getArch()));
143 TheTriple
.setTriple(Triple::normalize(TripleName
));
145 // Get the target specific parser.
147 const Target
*TheTarget
= TargetRegistry::lookupTarget(ArchName
, TheTriple
,
150 errs() << ToolName
<< ": " << Error
;
154 // Update the triple name and return the found target.
155 TripleName
= TheTriple
.getTriple();
159 void llvm::StringRefMemoryObject::anchor() { }
161 void llvm::DumpBytes(StringRef bytes
) {
162 static const char hex_rep
[] = "0123456789abcdef";
163 // FIXME: The real way to do this is to figure out the longest instruction
164 // and align to that size before printing. I'll fix this when I get
165 // around to outputting relocations.
166 // 15 is the longest x86 instruction
167 // 3 is for the hex rep of a byte + a space.
168 // 1 is for the null terminator.
169 enum { OutputSize
= (15 * 3) + 1 };
170 char output
[OutputSize
];
172 assert(bytes
.size() <= 15
173 && "DumpBytes only supports instructions of up to 15 bytes");
174 memset(output
, ' ', sizeof(output
));
176 for (StringRef::iterator i
= bytes
.begin(),
177 e
= bytes
.end(); i
!= e
; ++i
) {
178 output
[index
] = hex_rep
[(*i
& 0xF0) >> 4];
179 output
[index
+ 1] = hex_rep
[*i
& 0xF];
183 output
[sizeof(output
) - 1] = 0;
187 bool llvm::RelocAddressLess(RelocationRef a
, RelocationRef b
) {
188 uint64_t a_addr
, b_addr
;
189 if (error(a
.getAddress(a_addr
))) return false;
190 if (error(b
.getAddress(b_addr
))) return false;
191 return a_addr
< b_addr
;
194 static void DisassembleObject(const ObjectFile
*Obj
, bool InlineRelocs
) {
195 const Target
*TheTarget
= getTarget(Obj
);
196 // getTarget() will have already issued a diagnostic if necessary, so
197 // just bail here if it failed.
201 // Package up features to be passed to target/subtarget
202 std::string FeaturesStr
;
204 SubtargetFeatures Features
;
205 for (unsigned i
= 0; i
!= MAttrs
.size(); ++i
)
206 Features
.AddFeature(MAttrs
[i
]);
207 FeaturesStr
= Features
.getString();
211 for (section_iterator i
= Obj
->begin_sections(),
212 e
= Obj
->end_sections();
213 i
!= e
; i
.increment(ec
)) {
214 if (error(ec
)) break;
216 if (error(i
->isText(text
))) break;
219 uint64_t SectionAddr
;
220 if (error(i
->getAddress(SectionAddr
))) break;
222 // Make a list of all the symbols in this section.
223 std::vector
<std::pair
<uint64_t, StringRef
> > Symbols
;
224 for (symbol_iterator si
= Obj
->begin_symbols(),
225 se
= Obj
->end_symbols();
226 si
!= se
; si
.increment(ec
)) {
228 if (!error(i
->containsSymbol(*si
, contains
)) && contains
) {
230 if (error(si
->getAddress(Address
))) break;
231 Address
-= SectionAddr
;
234 if (error(si
->getName(Name
))) break;
235 Symbols
.push_back(std::make_pair(Address
, Name
));
239 // Sort the symbols by address, just in case they didn't come in that way.
240 array_pod_sort(Symbols
.begin(), Symbols
.end());
242 // Make a list of all the relocations for this section.
243 std::vector
<RelocationRef
> Rels
;
245 for (relocation_iterator ri
= i
->begin_relocations(),
246 re
= i
->end_relocations();
247 ri
!= re
; ri
.increment(ec
)) {
248 if (error(ec
)) break;
253 // Sort relocations by address.
254 std::sort(Rels
.begin(), Rels
.end(), RelocAddressLess
);
256 StringRef SegmentName
= "";
257 if (const MachOObjectFile
*MachO
= dyn_cast
<const MachOObjectFile
>(Obj
)) {
258 DataRefImpl DR
= i
->getRawDataRefImpl();
259 if (error(MachO
->getSectionFinalSegmentName(DR
, SegmentName
)))
263 if (error(i
->getName(name
))) break;
264 outs() << "Disassembly of section ";
265 if (!SegmentName
.empty())
266 outs() << SegmentName
<< ",";
267 outs() << name
<< ':';
269 // If the section has no symbols just insert a dummy one and disassemble
270 // the whole section.
272 Symbols
.push_back(std::make_pair(0, name
));
274 // Set up disassembler.
275 OwningPtr
<const MCAsmInfo
> AsmInfo(TheTarget
->createMCAsmInfo(TripleName
));
278 errs() << "error: no assembly info for target " << TripleName
<< "\n";
282 OwningPtr
<const MCSubtargetInfo
> STI(
283 TheTarget
->createMCSubtargetInfo(TripleName
, "", FeaturesStr
));
286 errs() << "error: no subtarget info for target " << TripleName
<< "\n";
290 OwningPtr
<const MCDisassembler
> DisAsm(
291 TheTarget
->createMCDisassembler(*STI
));
293 errs() << "error: no disassembler for target " << TripleName
<< "\n";
297 OwningPtr
<const MCRegisterInfo
> MRI(TheTarget
->createMCRegInfo(TripleName
));
299 errs() << "error: no register info for target " << TripleName
<< "\n";
303 OwningPtr
<const MCInstrInfo
> MII(TheTarget
->createMCInstrInfo());
305 errs() << "error: no instruction info for target " << TripleName
<< "\n";
309 int AsmPrinterVariant
= AsmInfo
->getAssemblerDialect();
310 OwningPtr
<MCInstPrinter
> IP(TheTarget
->createMCInstPrinter(
311 AsmPrinterVariant
, *AsmInfo
, *MII
, *MRI
, *STI
));
313 errs() << "error: no instruction printer for target " << TripleName
319 if (error(i
->getContents(Bytes
))) break;
320 StringRefMemoryObject
memoryObject(Bytes
);
324 if (error(i
->getSize(SectSize
))) break;
326 std::vector
<RelocationRef
>::const_iterator rel_cur
= Rels
.begin();
327 std::vector
<RelocationRef
>::const_iterator rel_end
= Rels
.end();
328 // Disassemble symbol by symbol.
329 for (unsigned si
= 0, se
= Symbols
.size(); si
!= se
; ++si
) {
330 uint64_t Start
= Symbols
[si
].first
;
332 // The end is either the size of the section or the beginning of the next
336 // Make sure this symbol takes up space.
337 else if (Symbols
[si
+ 1].first
!= Start
)
338 End
= Symbols
[si
+ 1].first
- 1;
340 // This symbol has the same address as the next symbol. Skip it.
343 outs() << '\n' << Symbols
[si
].second
<< ":\n";
346 raw_ostream
&DebugOut
= DebugFlag
? dbgs() : nulls();
348 raw_ostream
&DebugOut
= nulls();
351 for (Index
= Start
; Index
< End
; Index
+= Size
) {
354 if (DisAsm
->getInstruction(Inst
, Size
, memoryObject
, Index
,
355 DebugOut
, nulls())) {
356 outs() << format("%8" PRIx64
":", SectionAddr
+ Index
);
357 if (!NoShowRawInsn
) {
359 DumpBytes(StringRef(Bytes
.data() + Index
, Size
));
361 IP
->printInst(&Inst
, outs(), "");
364 errs() << ToolName
<< ": warning: invalid instruction encoding\n";
366 Size
= 1; // skip illegible bytes
369 // Print relocation for instruction.
370 while (rel_cur
!= rel_end
) {
373 SmallString
<16> name
;
376 // If this relocation is hidden, skip it.
377 if (error(rel_cur
->getHidden(hidden
))) goto skip_print_rel
;
378 if (hidden
) goto skip_print_rel
;
380 if (error(rel_cur
->getAddress(addr
))) goto skip_print_rel
;
381 // Stop when rel_cur's address is past the current instruction.
382 if (addr
>= Index
+ Size
) break;
383 if (error(rel_cur
->getTypeName(name
))) goto skip_print_rel
;
384 if (error(rel_cur
->getValueString(val
))) goto skip_print_rel
;
386 outs() << format("\t\t\t%8" PRIx64
": ", SectionAddr
+ addr
) << name
387 << "\t" << val
<< "\n";
397 static void PrintRelocations(const ObjectFile
*o
) {
399 for (section_iterator si
= o
->begin_sections(), se
= o
->end_sections();
400 si
!= se
; si
.increment(ec
)){
401 if (error(ec
)) return;
402 if (si
->begin_relocations() == si
->end_relocations())
405 if (error(si
->getName(secname
))) continue;
406 outs() << "RELOCATION RECORDS FOR [" << secname
<< "]:\n";
407 for (relocation_iterator ri
= si
->begin_relocations(),
408 re
= si
->end_relocations();
409 ri
!= re
; ri
.increment(ec
)) {
410 if (error(ec
)) return;
414 SmallString
<32> relocname
;
415 SmallString
<32> valuestr
;
416 if (error(ri
->getHidden(hidden
))) continue;
417 if (hidden
) continue;
418 if (error(ri
->getTypeName(relocname
))) continue;
419 if (error(ri
->getAddress(address
))) continue;
420 if (error(ri
->getValueString(valuestr
))) continue;
421 outs() << address
<< " " << relocname
<< " " << valuestr
<< "\n";
427 static void PrintSectionHeaders(const ObjectFile
*o
) {
428 outs() << "Sections:\n"
429 "Idx Name Size Address Type\n";
432 for (section_iterator si
= o
->begin_sections(), se
= o
->end_sections();
433 si
!= se
; si
.increment(ec
)) {
434 if (error(ec
)) return;
436 if (error(si
->getName(Name
))) return;
438 if (error(si
->getAddress(Address
))) return;
440 if (error(si
->getSize(Size
))) return;
441 bool Text
, Data
, BSS
;
442 if (error(si
->isText(Text
))) return;
443 if (error(si
->isData(Data
))) return;
444 if (error(si
->isBSS(BSS
))) return;
445 std::string Type
= (std::string(Text
? "TEXT " : "") +
446 (Data
? "DATA " : "") + (BSS
? "BSS" : ""));
447 outs() << format("%3d %-13s %08" PRIx64
" %016" PRIx64
" %s\n",
448 i
, Name
.str().c_str(), Size
, Address
, Type
.c_str());
453 static void PrintSectionContents(const ObjectFile
*o
) {
455 for (section_iterator si
= o
->begin_sections(),
456 se
= o
->end_sections();
457 si
!= se
; si
.increment(ec
)) {
458 if (error(ec
)) return;
462 if (error(si
->getName(Name
))) continue;
463 if (error(si
->getContents(Contents
))) continue;
464 if (error(si
->getAddress(BaseAddr
))) continue;
466 outs() << "Contents of section " << Name
<< ":\n";
468 // Dump out the content as hex and printable ascii characters.
469 for (std::size_t addr
= 0, end
= Contents
.size(); addr
< end
; addr
+= 16) {
470 outs() << format(" %04" PRIx64
" ", BaseAddr
+ addr
);
472 for (std::size_t i
= 0; i
< 16; ++i
) {
473 if (i
!= 0 && i
% 4 == 0)
476 outs() << hexdigit((Contents
[addr
+ i
] >> 4) & 0xF, true)
477 << hexdigit(Contents
[addr
+ i
] & 0xF, true);
483 for (std::size_t i
= 0; i
< 16 && addr
+ i
< end
; ++i
) {
484 if (std::isprint(static_cast<unsigned char>(Contents
[addr
+ i
]) & 0xFF))
485 outs() << Contents
[addr
+ i
];
494 static void PrintCOFFSymbolTable(const COFFObjectFile
*coff
) {
495 const coff_file_header
*header
;
496 if (error(coff
->getHeader(header
))) return;
498 const coff_symbol
*symbol
= 0;
499 for (int i
= 0, e
= header
->NumberOfSymbols
; i
!= e
; ++i
) {
501 // Figure out which type of aux this is.
502 if (symbol
->StorageClass
== COFF::IMAGE_SYM_CLASS_STATIC
503 && symbol
->Value
== 0) { // Section definition.
504 const coff_aux_section_definition
*asd
;
505 if (error(coff
->getAuxSymbol
<coff_aux_section_definition
>(i
, asd
)))
508 << format("scnlen 0x%x nreloc %d nlnno %d checksum 0x%x "
509 , unsigned(asd
->Length
)
510 , unsigned(asd
->NumberOfRelocations
)
511 , unsigned(asd
->NumberOfLinenumbers
)
512 , unsigned(asd
->CheckSum
))
513 << format("assoc %d comdat %d\n"
514 , unsigned(asd
->Number
)
515 , unsigned(asd
->Selection
));
517 outs() << "AUX Unknown\n";
520 if (error(coff
->getSymbol(i
, symbol
))) return;
521 if (error(coff
->getSymbolName(symbol
, name
))) return;
522 outs() << "[" << format("%2d", i
) << "]"
523 << "(sec " << format("%2d", int(symbol
->SectionNumber
)) << ")"
524 << "(fl 0x00)" // Flag bits, which COFF doesn't have.
525 << "(ty " << format("%3x", unsigned(symbol
->Type
)) << ")"
526 << "(scl " << format("%3x", unsigned(symbol
->StorageClass
)) << ") "
527 << "(nx " << unsigned(symbol
->NumberOfAuxSymbols
) << ") "
528 << "0x" << format("%08x", unsigned(symbol
->Value
)) << " "
530 aux_count
= symbol
->NumberOfAuxSymbols
;
535 static void PrintSymbolTable(const ObjectFile
*o
) {
536 outs() << "SYMBOL TABLE:\n";
538 if (const COFFObjectFile
*coff
= dyn_cast
<const COFFObjectFile
>(o
))
539 PrintCOFFSymbolTable(coff
);
542 for (symbol_iterator si
= o
->begin_symbols(),
543 se
= o
->end_symbols(); si
!= se
; si
.increment(ec
)) {
544 if (error(ec
)) return;
547 SymbolRef::Type Type
;
550 section_iterator Section
= o
->end_sections();
551 if (error(si
->getName(Name
))) continue;
552 if (error(si
->getAddress(Address
))) continue;
553 if (error(si
->getFlags(Flags
))) continue;
554 if (error(si
->getType(Type
))) continue;
555 if (error(si
->getSize(Size
))) continue;
556 if (error(si
->getSection(Section
))) continue;
558 bool Global
= Flags
& SymbolRef::SF_Global
;
559 bool Weak
= Flags
& SymbolRef::SF_Weak
;
560 bool Absolute
= Flags
& SymbolRef::SF_Absolute
;
562 if (Address
== UnknownAddressOrSize
)
564 if (Size
== UnknownAddressOrSize
)
567 if (Type
!= SymbolRef::ST_Unknown
)
568 GlobLoc
= Global
? 'g' : 'l';
569 char Debug
= (Type
== SymbolRef::ST_Debug
|| Type
== SymbolRef::ST_File
)
572 if (Type
== SymbolRef::ST_File
)
574 else if (Type
== SymbolRef::ST_Function
)
577 const char *Fmt
= o
->getBytesInAddress() > 4 ? "%016" PRIx64
:
580 outs() << format(Fmt
, Address
) << " "
581 << GlobLoc
// Local -> 'l', Global -> 'g', Neither -> ' '
582 << (Weak
? 'w' : ' ') // Weak?
583 << ' ' // Constructor. Not supported yet.
584 << ' ' // Warning. Not supported yet.
585 << ' ' // Indirect reference to another symbol.
586 << Debug
// Debugging (d) or dynamic (D) symbol.
587 << FileFunc
// Name of function (F), file (f) or object (O).
591 else if (Section
== o
->end_sections())
594 if (const MachOObjectFile
*MachO
= dyn_cast
<const MachOObjectFile
>(o
)) {
595 StringRef SegmentName
;
596 DataRefImpl DR
= Section
->getRawDataRefImpl();
597 if (error(MachO
->getSectionFinalSegmentName(DR
, SegmentName
)))
599 outs() << SegmentName
<< ",";
601 StringRef SectionName
;
602 if (error(Section
->getName(SectionName
)))
604 outs() << SectionName
;
607 << format("%08" PRIx64
" ", Size
)
614 static void PrintUnwindInfo(const ObjectFile
*o
) {
615 outs() << "Unwind info:\n\n";
617 if (const COFFObjectFile
*coff
= dyn_cast
<COFFObjectFile
>(o
)) {
618 printCOFFUnwindInfo(coff
);
620 // TODO: Extract DWARF dump tool to objdump.
621 errs() << "This operation is only currently supported "
622 "for COFF object files.\n";
627 static void DumpObject(const ObjectFile
*o
) {
629 outs() << o
->getFileName()
630 << ":\tfile format " << o
->getFileFormatName() << "\n\n";
633 DisassembleObject(o
, Relocations
);
634 if (Relocations
&& !Disassemble
)
637 PrintSectionHeaders(o
);
639 PrintSectionContents(o
);
644 if (PrivateHeaders
&& o
->isELF())
645 printELFFileHeader(o
);
648 /// @brief Dump each object file in \a a;
649 static void DumpArchive(const Archive
*a
) {
650 for (Archive::child_iterator i
= a
->begin_children(),
651 e
= a
->end_children(); i
!= e
; ++i
) {
652 OwningPtr
<Binary
> child
;
653 if (error_code ec
= i
->getAsBinary(child
)) {
654 // Ignore non-object files.
655 if (ec
!= object_error::invalid_file_type
)
656 errs() << ToolName
<< ": '" << a
->getFileName() << "': " << ec
.message()
660 if (ObjectFile
*o
= dyn_cast
<ObjectFile
>(child
.get()))
663 errs() << ToolName
<< ": '" << a
->getFileName() << "': "
664 << "Unrecognized file type.\n";
668 /// @brief Open file and figure out how to dump it.
669 static void DumpInput(StringRef file
) {
670 // If file isn't stdin, check that it exists.
671 if (file
!= "-" && !sys::fs::exists(file
)) {
672 errs() << ToolName
<< ": '" << file
<< "': " << "No such file\n";
676 if (MachOOpt
&& Disassemble
) {
677 DisassembleInputMachO(file
);
681 // Attempt to open the binary.
682 OwningPtr
<Binary
> binary
;
683 if (error_code ec
= createBinary(file
, binary
)) {
684 errs() << ToolName
<< ": '" << file
<< "': " << ec
.message() << ".\n";
688 if (Archive
*a
= dyn_cast
<Archive
>(binary
.get()))
690 else if (ObjectFile
*o
= dyn_cast
<ObjectFile
>(binary
.get()))
693 errs() << ToolName
<< ": '" << file
<< "': " << "Unrecognized file type.\n";
696 int main(int argc
, char **argv
) {
697 // Print a stack trace if we signal out.
698 sys::PrintStackTraceOnErrorSignal();
699 PrettyStackTraceProgram
X(argc
, argv
);
700 llvm_shutdown_obj Y
; // Call llvm_shutdown() on exit.
702 // Initialize targets and assembly printers/parsers.
703 llvm::InitializeAllTargetInfos();
704 llvm::InitializeAllTargetMCs();
705 llvm::InitializeAllAsmParsers();
706 llvm::InitializeAllDisassemblers();
708 // Register the target printer for --version.
709 cl::AddExtraVersionPrinter(TargetRegistry::printRegisteredTargetsForVersion
);
711 cl::ParseCommandLineOptions(argc
, argv
, "llvm object file dumper\n");
712 TripleName
= Triple::normalize(TripleName
);
716 // Defaults to a.out if no filenames specified.
717 if (InputFilenames
.size() == 0)
718 InputFilenames
.push_back("a.out");
726 && !PrivateHeaders
) {
727 cl::PrintHelpMessage();
731 std::for_each(InputFilenames
.begin(), InputFilenames
.end(),