1 //===- llvm-cxxdump.cpp - Dump C++ data in an Object File -------*- C++ -*-===//
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 // Dumps C++ data resident in object files and archives.
12 //===----------------------------------------------------------------------===//
14 #include "llvm-cxxdump.h"
16 #include "llvm/ADT/ArrayRef.h"
17 #include "llvm/Object/Archive.h"
18 #include "llvm/Object/ObjectFile.h"
19 #include "llvm/Object/SymbolSize.h"
20 #include "llvm/Support/Debug.h"
21 #include "llvm/Support/Endian.h"
22 #include "llvm/Support/FileSystem.h"
23 #include "llvm/Support/ManagedStatic.h"
24 #include "llvm/Support/PrettyStackTrace.h"
25 #include "llvm/Support/Signals.h"
26 #include "llvm/Support/TargetRegistry.h"
27 #include "llvm/Support/TargetSelect.h"
28 #include "llvm/Support/raw_ostream.h"
31 #include <system_error>
34 using namespace llvm::object;
35 using namespace llvm::support;
38 cl::list<std::string> InputFilenames(cl::Positional,
39 cl::desc("<input object files>"),
43 static int ReturnValue = EXIT_SUCCESS;
47 static bool error(std::error_code EC) {
51 ReturnValue = EXIT_FAILURE;
52 outs() << "\nError reading file: " << EC.message() << ".\n";
59 static void reportError(StringRef Input, StringRef Message) {
63 errs() << Input << ": " << Message << "\n";
65 ReturnValue = EXIT_FAILURE;
68 static void reportError(StringRef Input, std::error_code EC) {
69 reportError(Input, EC.message());
72 static SmallVectorImpl<SectionRef> &getRelocSections(const ObjectFile *Obj,
73 const SectionRef &Sec) {
74 static bool MappingDone = false;
75 static std::map<SectionRef, SmallVector<SectionRef, 1>> SectionRelocMap;
77 for (const SectionRef &Section : Obj->sections()) {
78 section_iterator Sec2 = Section.getRelocatedSection();
79 if (Sec2 != Obj->section_end())
80 SectionRelocMap[*Sec2].push_back(Section);
84 return SectionRelocMap[Sec];
87 static bool collectRelocatedSymbols(const ObjectFile *Obj,
88 const SectionRef &Sec, uint64_t SecAddress,
89 uint64_t SymAddress, uint64_t SymSize,
90 StringRef *I, StringRef *E) {
91 uint64_t SymOffset = SymAddress - SecAddress;
92 uint64_t SymEnd = SymOffset + SymSize;
93 for (const SectionRef &SR : getRelocSections(Obj, Sec)) {
94 for (const object::RelocationRef &Reloc : SR.relocations()) {
97 const object::symbol_iterator RelocSymI = Reloc.getSymbol();
98 if (RelocSymI == Obj->symbol_end())
100 ErrorOr<StringRef> RelocSymName = RelocSymI->getName();
101 if (error(RelocSymName.getError()))
103 uint64_t Offset = Reloc.getOffset();
104 if (Offset >= SymOffset && Offset < SymEnd) {
113 static bool collectRelocationOffsets(
114 const ObjectFile *Obj, const SectionRef &Sec, uint64_t SecAddress,
115 uint64_t SymAddress, uint64_t SymSize, StringRef SymName,
116 std::map<std::pair<StringRef, uint64_t>, StringRef> &Collection) {
117 uint64_t SymOffset = SymAddress - SecAddress;
118 uint64_t SymEnd = SymOffset + SymSize;
119 for (const SectionRef &SR : getRelocSections(Obj, Sec)) {
120 for (const object::RelocationRef &Reloc : SR.relocations()) {
121 const object::symbol_iterator RelocSymI = Reloc.getSymbol();
122 if (RelocSymI == Obj->symbol_end())
124 ErrorOr<StringRef> RelocSymName = RelocSymI->getName();
125 if (error(RelocSymName.getError()))
127 uint64_t Offset = Reloc.getOffset();
128 if (Offset >= SymOffset && Offset < SymEnd)
129 Collection[std::make_pair(SymName, Offset - SymOffset)] = *RelocSymName;
135 static void dumpCXXData(const ObjectFile *Obj) {
136 struct CompleteObjectLocator {
137 StringRef Symbols[2];
138 ArrayRef<little32_t> Data;
140 struct ClassHierarchyDescriptor {
141 StringRef Symbols[1];
142 ArrayRef<little32_t> Data;
144 struct BaseClassDescriptor {
145 StringRef Symbols[2];
146 ArrayRef<little32_t> Data;
148 struct TypeDescriptor {
149 StringRef Symbols[1];
151 StringRef MangledName;
156 struct CatchableTypeArray {
159 struct CatchableType {
161 uint32_t NonVirtualBaseAdjustmentOffset;
162 int32_t VirtualBasePointerOffset;
163 uint32_t VirtualBaseAdjustmentOffset;
165 StringRef Symbols[2];
167 std::map<std::pair<StringRef, uint64_t>, StringRef> VFTableEntries;
168 std::map<std::pair<StringRef, uint64_t>, StringRef> TIEntries;
169 std::map<std::pair<StringRef, uint64_t>, StringRef> CTAEntries;
170 std::map<StringRef, ArrayRef<little32_t>> VBTables;
171 std::map<StringRef, CompleteObjectLocator> COLs;
172 std::map<StringRef, ClassHierarchyDescriptor> CHDs;
173 std::map<std::pair<StringRef, uint64_t>, StringRef> BCAEntries;
174 std::map<StringRef, BaseClassDescriptor> BCDs;
175 std::map<StringRef, TypeDescriptor> TDs;
176 std::map<StringRef, ThrowInfo> TIs;
177 std::map<StringRef, CatchableTypeArray> CTAs;
178 std::map<StringRef, CatchableType> CTs;
180 std::map<std::pair<StringRef, uint64_t>, StringRef> VTableSymEntries;
181 std::map<std::pair<StringRef, uint64_t>, int64_t> VTableDataEntries;
182 std::map<std::pair<StringRef, uint64_t>, StringRef> VTTEntries;
183 std::map<StringRef, StringRef> TINames;
185 uint8_t BytesInAddress = Obj->getBytesInAddress();
187 std::vector<std::pair<SymbolRef, uint64_t>> SymAddr =
188 object::computeSymbolSizes(*Obj);
190 for (auto &P : SymAddr) {
191 object::SymbolRef Sym = P.first;
192 uint64_t SymSize = P.second;
193 ErrorOr<StringRef> SymNameOrErr = Sym.getName();
194 if (error(SymNameOrErr.getError()))
196 StringRef SymName = *SymNameOrErr;
197 object::section_iterator SecI(Obj->section_begin());
198 if (error(Sym.getSection(SecI)))
200 // Skip external symbols.
201 if (SecI == Obj->section_end())
203 const SectionRef &Sec = *SecI;
204 // Skip virtual or BSS sections.
205 if (Sec.isBSS() || Sec.isVirtual())
207 StringRef SecContents;
208 if (error(Sec.getContents(SecContents)))
210 ErrorOr<uint64_t> SymAddressOrErr = Sym.getAddress();
211 if (error(SymAddressOrErr.getError()))
213 uint64_t SymAddress = *SymAddressOrErr;
214 uint64_t SecAddress = Sec.getAddress();
215 uint64_t SecSize = Sec.getSize();
216 uint64_t SymOffset = SymAddress - SecAddress;
217 StringRef SymContents = SecContents.substr(SymOffset, SymSize);
219 // VFTables in the MS-ABI start with '??_7' and are contained within their
220 // own COMDAT section. We then determine the contents of the VFTable by
221 // looking at each relocation in the section.
222 if (SymName.startswith("??_7")) {
223 // Each relocation either names a virtual method or a thunk. We note the
224 // offset into the section and the symbol used for the relocation.
225 collectRelocationOffsets(Obj, Sec, SecAddress, SecAddress, SecSize,
226 SymName, VFTableEntries);
228 // VBTables in the MS-ABI start with '??_8' and are filled with 32-bit
229 // offsets of virtual bases.
230 else if (SymName.startswith("??_8")) {
231 ArrayRef<little32_t> VBTableData(
232 reinterpret_cast<const little32_t *>(SymContents.data()),
233 SymContents.size() / sizeof(little32_t));
234 VBTables[SymName] = VBTableData;
236 // Complete object locators in the MS-ABI start with '??_R4'
237 else if (SymName.startswith("??_R4")) {
238 CompleteObjectLocator COL;
239 COL.Data = ArrayRef<little32_t>(
240 reinterpret_cast<const little32_t *>(SymContents.data()), 3);
241 StringRef *I = std::begin(COL.Symbols), *E = std::end(COL.Symbols);
242 if (collectRelocatedSymbols(Obj, Sec, SecAddress, SymAddress, SymSize, I,
247 // Class hierarchy descriptors in the MS-ABI start with '??_R3'
248 else if (SymName.startswith("??_R3")) {
249 ClassHierarchyDescriptor CHD;
250 CHD.Data = ArrayRef<little32_t>(
251 reinterpret_cast<const little32_t *>(SymContents.data()), 3);
252 StringRef *I = std::begin(CHD.Symbols), *E = std::end(CHD.Symbols);
253 if (collectRelocatedSymbols(Obj, Sec, SecAddress, SymAddress, SymSize, I,
258 // Class hierarchy descriptors in the MS-ABI start with '??_R2'
259 else if (SymName.startswith("??_R2")) {
260 // Each relocation names a base class descriptor. We note the offset into
261 // the section and the symbol used for the relocation.
262 collectRelocationOffsets(Obj, Sec, SecAddress, SymAddress, SymSize,
263 SymName, BCAEntries);
265 // Base class descriptors in the MS-ABI start with '??_R1'
266 else if (SymName.startswith("??_R1")) {
267 BaseClassDescriptor BCD;
268 BCD.Data = ArrayRef<little32_t>(
269 reinterpret_cast<const little32_t *>(SymContents.data()) + 1, 5);
270 StringRef *I = std::begin(BCD.Symbols), *E = std::end(BCD.Symbols);
271 if (collectRelocatedSymbols(Obj, Sec, SecAddress, SymAddress, SymSize, I,
276 // Type descriptors in the MS-ABI start with '??_R0'
277 else if (SymName.startswith("??_R0")) {
278 const char *DataPtr = SymContents.drop_front(BytesInAddress).data();
280 if (BytesInAddress == 8)
281 TD.AlwaysZero = *reinterpret_cast<const little64_t *>(DataPtr);
283 TD.AlwaysZero = *reinterpret_cast<const little32_t *>(DataPtr);
284 TD.MangledName = SymContents.drop_front(BytesInAddress * 2);
285 StringRef *I = std::begin(TD.Symbols), *E = std::end(TD.Symbols);
286 if (collectRelocatedSymbols(Obj, Sec, SecAddress, SymAddress, SymSize, I,
291 // Throw descriptors in the MS-ABI start with '_TI'
292 else if (SymName.startswith("_TI") || SymName.startswith("__TI")) {
294 TI.Flags = *reinterpret_cast<const little32_t *>(SymContents.data());
295 collectRelocationOffsets(Obj, Sec, SecAddress, SymAddress, SymSize,
299 // Catchable type arrays in the MS-ABI start with _CTA or __CTA.
300 else if (SymName.startswith("_CTA") || SymName.startswith("__CTA")) {
301 CatchableTypeArray CTA;
303 *reinterpret_cast<const little32_t *>(SymContents.data());
304 collectRelocationOffsets(Obj, Sec, SecAddress, SymAddress, SymSize,
305 SymName, CTAEntries);
308 // Catchable types in the MS-ABI start with _CT or __CT.
309 else if (SymName.startswith("_CT") || SymName.startswith("__CT")) {
310 const little32_t *DataPtr =
311 reinterpret_cast<const little32_t *>(SymContents.data());
313 CT.Flags = DataPtr[0];
314 CT.NonVirtualBaseAdjustmentOffset = DataPtr[2];
315 CT.VirtualBasePointerOffset = DataPtr[3];
316 CT.VirtualBaseAdjustmentOffset = DataPtr[4];
317 CT.Size = DataPtr[5];
318 StringRef *I = std::begin(CT.Symbols), *E = std::end(CT.Symbols);
319 if (collectRelocatedSymbols(Obj, Sec, SecAddress, SymAddress, SymSize, I,
324 // Construction vtables in the Itanium ABI start with '_ZTT' or '__ZTT'.
325 else if (SymName.startswith("_ZTT") || SymName.startswith("__ZTT")) {
326 collectRelocationOffsets(Obj, Sec, SecAddress, SymAddress, SymSize,
327 SymName, VTTEntries);
329 // Typeinfo names in the Itanium ABI start with '_ZTS' or '__ZTS'.
330 else if (SymName.startswith("_ZTS") || SymName.startswith("__ZTS")) {
331 TINames[SymName] = SymContents.slice(0, SymContents.find('\0'));
333 // Vtables in the Itanium ABI start with '_ZTV' or '__ZTV'.
334 else if (SymName.startswith("_ZTV") || SymName.startswith("__ZTV")) {
335 collectRelocationOffsets(Obj, Sec, SecAddress, SymAddress, SymSize,
336 SymName, VTableSymEntries);
337 for (uint64_t SymOffI = 0; SymOffI < SymSize; SymOffI += BytesInAddress) {
338 auto Key = std::make_pair(SymName, SymOffI);
339 if (VTableSymEntries.count(Key))
341 const char *DataPtr =
342 SymContents.substr(SymOffI, BytesInAddress).data();
344 if (BytesInAddress == 8)
345 VData = *reinterpret_cast<const little64_t *>(DataPtr);
347 VData = *reinterpret_cast<const little32_t *>(DataPtr);
348 VTableDataEntries[Key] = VData;
351 // Typeinfo structures in the Itanium ABI start with '_ZTI' or '__ZTI'.
352 else if (SymName.startswith("_ZTI") || SymName.startswith("__ZTI")) {
353 // FIXME: Do something with these!
356 for (const auto &VFTableEntry : VFTableEntries) {
357 StringRef VFTableName = VFTableEntry.first.first;
358 uint64_t Offset = VFTableEntry.first.second;
359 StringRef SymName = VFTableEntry.second;
360 outs() << VFTableName << '[' << Offset << "]: " << SymName << '\n';
362 for (const auto &VBTable : VBTables) {
363 StringRef VBTableName = VBTable.first;
365 for (little32_t Offset : VBTable.second) {
366 outs() << VBTableName << '[' << Idx << "]: " << Offset << '\n';
367 Idx += sizeof(Offset);
370 for (const auto &COLPair : COLs) {
371 StringRef COLName = COLPair.first;
372 const CompleteObjectLocator &COL = COLPair.second;
373 outs() << COLName << "[IsImageRelative]: " << COL.Data[0] << '\n';
374 outs() << COLName << "[OffsetToTop]: " << COL.Data[1] << '\n';
375 outs() << COLName << "[VFPtrOffset]: " << COL.Data[2] << '\n';
376 outs() << COLName << "[TypeDescriptor]: " << COL.Symbols[0] << '\n';
377 outs() << COLName << "[ClassHierarchyDescriptor]: " << COL.Symbols[1]
380 for (const auto &CHDPair : CHDs) {
381 StringRef CHDName = CHDPair.first;
382 const ClassHierarchyDescriptor &CHD = CHDPair.second;
383 outs() << CHDName << "[AlwaysZero]: " << CHD.Data[0] << '\n';
384 outs() << CHDName << "[Flags]: " << CHD.Data[1] << '\n';
385 outs() << CHDName << "[NumClasses]: " << CHD.Data[2] << '\n';
386 outs() << CHDName << "[BaseClassArray]: " << CHD.Symbols[0] << '\n';
388 for (const auto &BCAEntry : BCAEntries) {
389 StringRef BCAName = BCAEntry.first.first;
390 uint64_t Offset = BCAEntry.first.second;
391 StringRef SymName = BCAEntry.second;
392 outs() << BCAName << '[' << Offset << "]: " << SymName << '\n';
394 for (const auto &BCDPair : BCDs) {
395 StringRef BCDName = BCDPair.first;
396 const BaseClassDescriptor &BCD = BCDPair.second;
397 outs() << BCDName << "[TypeDescriptor]: " << BCD.Symbols[0] << '\n';
398 outs() << BCDName << "[NumBases]: " << BCD.Data[0] << '\n';
399 outs() << BCDName << "[OffsetInVBase]: " << BCD.Data[1] << '\n';
400 outs() << BCDName << "[VBPtrOffset]: " << BCD.Data[2] << '\n';
401 outs() << BCDName << "[OffsetInVBTable]: " << BCD.Data[3] << '\n';
402 outs() << BCDName << "[Flags]: " << BCD.Data[4] << '\n';
403 outs() << BCDName << "[ClassHierarchyDescriptor]: " << BCD.Symbols[1]
406 for (const auto &TDPair : TDs) {
407 StringRef TDName = TDPair.first;
408 const TypeDescriptor &TD = TDPair.second;
409 outs() << TDName << "[VFPtr]: " << TD.Symbols[0] << '\n';
410 outs() << TDName << "[AlwaysZero]: " << TD.AlwaysZero << '\n';
411 outs() << TDName << "[MangledName]: ";
412 outs().write_escaped(TD.MangledName.rtrim(StringRef("\0", 1)),
413 /*UseHexEscapes=*/true)
416 for (const auto &TIPair : TIs) {
417 StringRef TIName = TIPair.first;
418 const ThrowInfo &TI = TIPair.second;
419 auto dumpThrowInfoFlag = [&](const char *Name, uint32_t Flag) {
420 outs() << TIName << "[Flags." << Name
421 << "]: " << (TI.Flags & Flag ? "true" : "false") << '\n';
423 auto dumpThrowInfoSymbol = [&](const char *Name, int Offset) {
424 outs() << TIName << '[' << Name << "]: ";
425 auto Entry = TIEntries.find(std::make_pair(TIName, Offset));
426 outs() << (Entry == TIEntries.end() ? "null" : Entry->second) << '\n';
428 outs() << TIName << "[Flags]: " << TI.Flags << '\n';
429 dumpThrowInfoFlag("Const", 1);
430 dumpThrowInfoFlag("Volatile", 2);
431 dumpThrowInfoSymbol("CleanupFn", 4);
432 dumpThrowInfoSymbol("ForwardCompat", 8);
433 dumpThrowInfoSymbol("CatchableTypeArray", 12);
435 for (const auto &CTAPair : CTAs) {
436 StringRef CTAName = CTAPair.first;
437 const CatchableTypeArray &CTA = CTAPair.second;
439 outs() << CTAName << "[NumEntries]: " << CTA.NumEntries << '\n';
442 for (auto I = CTAEntries.lower_bound(std::make_pair(CTAName, 0)),
443 E = CTAEntries.upper_bound(std::make_pair(CTAName, UINT64_MAX));
445 outs() << CTAName << '[' << Idx++ << "]: " << I->second << '\n';
447 for (const auto &CTPair : CTs) {
448 StringRef CTName = CTPair.first;
449 const CatchableType &CT = CTPair.second;
450 auto dumpCatchableTypeFlag = [&](const char *Name, uint32_t Flag) {
451 outs() << CTName << "[Flags." << Name
452 << "]: " << (CT.Flags & Flag ? "true" : "false") << '\n';
454 outs() << CTName << "[Flags]: " << CT.Flags << '\n';
455 dumpCatchableTypeFlag("ScalarType", 1);
456 dumpCatchableTypeFlag("VirtualInheritance", 4);
457 outs() << CTName << "[TypeDescriptor]: " << CT.Symbols[0] << '\n';
458 outs() << CTName << "[NonVirtualBaseAdjustmentOffset]: "
459 << CT.NonVirtualBaseAdjustmentOffset << '\n';
461 << "[VirtualBasePointerOffset]: " << CT.VirtualBasePointerOffset
463 outs() << CTName << "[VirtualBaseAdjustmentOffset]: "
464 << CT.VirtualBaseAdjustmentOffset << '\n';
465 outs() << CTName << "[Size]: " << CT.Size << '\n';
467 << "[CopyCtor]: " << (CT.Symbols[1].empty() ? "null" : CT.Symbols[1])
470 for (const auto &VTTPair : VTTEntries) {
471 StringRef VTTName = VTTPair.first.first;
472 uint64_t VTTOffset = VTTPair.first.second;
473 StringRef VTTEntry = VTTPair.second;
474 outs() << VTTName << '[' << VTTOffset << "]: " << VTTEntry << '\n';
476 for (const auto &TIPair : TINames) {
477 StringRef TIName = TIPair.first;
478 outs() << TIName << ": " << TIPair.second << '\n';
480 auto VTableSymI = VTableSymEntries.begin();
481 auto VTableSymE = VTableSymEntries.end();
482 auto VTableDataI = VTableDataEntries.begin();
483 auto VTableDataE = VTableDataEntries.end();
485 bool SymDone = VTableSymI == VTableSymE;
486 bool DataDone = VTableDataI == VTableDataE;
487 if (SymDone && DataDone)
489 if (!SymDone && (DataDone || VTableSymI->first < VTableDataI->first)) {
490 StringRef VTableName = VTableSymI->first.first;
491 uint64_t Offset = VTableSymI->first.second;
492 StringRef VTableEntry = VTableSymI->second;
493 outs() << VTableName << '[' << Offset << "]: ";
494 outs() << VTableEntry;
499 if (!DataDone && (SymDone || VTableDataI->first < VTableSymI->first)) {
500 StringRef VTableName = VTableDataI->first.first;
501 uint64_t Offset = VTableDataI->first.second;
502 int64_t VTableEntry = VTableDataI->second;
503 outs() << VTableName << '[' << Offset << "]: ";
504 outs() << VTableEntry;
512 static void dumpArchive(const Archive *Arc) {
513 for (const Archive::Child &ArcC : Arc->children()) {
514 ErrorOr<std::unique_ptr<Binary>> ChildOrErr = ArcC.getAsBinary();
515 if (std::error_code EC = ChildOrErr.getError()) {
516 // Ignore non-object files.
517 if (EC != object_error::invalid_file_type)
518 reportError(Arc->getFileName(), EC.message());
522 if (ObjectFile *Obj = dyn_cast<ObjectFile>(&*ChildOrErr.get()))
525 reportError(Arc->getFileName(), cxxdump_error::unrecognized_file_format);
529 static void dumpInput(StringRef File) {
530 // If file isn't stdin, check that it exists.
531 if (File != "-" && !sys::fs::exists(File)) {
532 reportError(File, cxxdump_error::file_not_found);
536 // Attempt to open the binary.
537 ErrorOr<OwningBinary<Binary>> BinaryOrErr = createBinary(File);
538 if (std::error_code EC = BinaryOrErr.getError()) {
539 reportError(File, EC);
542 Binary &Binary = *BinaryOrErr.get().getBinary();
544 if (Archive *Arc = dyn_cast<Archive>(&Binary))
546 else if (ObjectFile *Obj = dyn_cast<ObjectFile>(&Binary))
549 reportError(File, cxxdump_error::unrecognized_file_format);
552 int main(int argc, const char *argv[]) {
553 sys::PrintStackTraceOnErrorSignal();
554 PrettyStackTraceProgram X(argc, argv);
557 // Initialize targets.
558 llvm::InitializeAllTargetInfos();
560 // Register the target printer for --version.
561 cl::AddExtraVersionPrinter(TargetRegistry::printRegisteredTargetsForVersion);
563 cl::ParseCommandLineOptions(argc, argv, "LLVM C++ ABI Data Dumper\n");
565 // Default to stdin if no filename is specified.
566 if (opts::InputFilenames.size() == 0)
567 opts::InputFilenames.push_back("-");
569 std::for_each(opts::InputFilenames.begin(), opts::InputFilenames.end(),