#include <thread>
#include <unordered_set>
#include "llvm/DebugInfo/DIContext.h"
#include "llvm/DebugInfo/DWARF/DWARFCompileUnit.h"
#include "llvm/DebugInfo/DWARF/DWARFContext.h"
#include "llvm/Support/Error.h"
#include "llvm/Support/ThreadPool.h"
#include "llvm/Support/raw_ostream.h"
#include "llvm/DebugInfo/GSYM/DwarfTransformer.h"
#include "llvm/DebugInfo/GSYM/FunctionInfo.h"
#include "llvm/DebugInfo/GSYM/GsymCreator.h"
#include "llvm/DebugInfo/GSYM/GsymReader.h"
#include "llvm/DebugInfo/GSYM/InlineInfo.h"
using namespace llvm;
using namespace gsym;
struct llvm::gsym::CUInfo {
const DWARFDebugLine::LineTable *LineTable;
const char *CompDir;
std::vector<uint32_t> FileCache;
uint64_t Language = 0;
uint8_t AddrSize = 0;
CUInfo(DWARFContext &DICtx, DWARFCompileUnit *CU) {
LineTable = DICtx.getLineTableForUnit(CU);
CompDir = CU->getCompilationDir();
FileCache.clear();
if (LineTable)
FileCache.assign(LineTable->Prologue.FileNames.size() + 1, UINT32_MAX);
DWARFDie Die = CU->getUnitDIE();
Language = dwarf::toUnsigned(Die.find(dwarf::DW_AT_language), 0);
AddrSize = CU->getAddressByteSize();
}
bool isHighestAddress(uint64_t Addr) const {
if (AddrSize == 4)
return Addr == UINT32_MAX;
else if (AddrSize == 8)
return Addr == UINT64_MAX;
return false;
}
uint32_t DWARFToGSYMFileIndex(GsymCreator &Gsym, uint32_t DwarfFileIdx) {
if (!LineTable)
return 0;
assert(DwarfFileIdx < FileCache.size());
uint32_t &GsymFileIdx = FileCache[DwarfFileIdx];
if (GsymFileIdx != UINT32_MAX)
return GsymFileIdx;
std::string File;
if (LineTable->getFileNameByIndex(
DwarfFileIdx, CompDir,
DILineInfoSpecifier::FileLineInfoKind::AbsoluteFilePath, File))
GsymFileIdx = Gsym.insertFile(File);
else
GsymFileIdx = 0;
return GsymFileIdx;
}
};
static DWARFDie GetParentDeclContextDIE(DWARFDie &Die) {
if (DWARFDie SpecDie =
Die.getAttributeValueAsReferencedDie(dwarf::DW_AT_specification)) {
if (DWARFDie SpecParent = GetParentDeclContextDIE(SpecDie))
return SpecParent;
}
if (DWARFDie AbstDie =
Die.getAttributeValueAsReferencedDie(dwarf::DW_AT_abstract_origin)) {
if (DWARFDie AbstParent = GetParentDeclContextDIE(AbstDie))
return AbstParent;
}
if (Die.getTag() == dwarf::DW_TAG_inlined_subroutine)
return DWARFDie();
DWARFDie ParentDie = Die.getParent();
if (!ParentDie)
return DWARFDie();
switch (ParentDie.getTag()) {
case dwarf::DW_TAG_namespace:
case dwarf::DW_TAG_structure_type:
case dwarf::DW_TAG_union_type:
case dwarf::DW_TAG_class_type:
case dwarf::DW_TAG_subprogram:
return ParentDie; case dwarf::DW_TAG_lexical_block:
return GetParentDeclContextDIE(ParentDie);
default:
break;
}
return DWARFDie();
}
static Optional<uint32_t> getQualifiedNameIndex(DWARFDie &Die,
uint64_t Language,
GsymCreator &Gsym) {
if (auto LinkageName =
dwarf::toString(Die.findRecursively({dwarf::DW_AT_MIPS_linkage_name,
dwarf::DW_AT_linkage_name}),
nullptr))
return Gsym.insertString(LinkageName, false);
StringRef ShortName(Die.getName(DINameKind::ShortName));
if (ShortName.empty())
return llvm::None;
if (!(Language == dwarf::DW_LANG_C_plus_plus ||
Language == dwarf::DW_LANG_C_plus_plus_03 ||
Language == dwarf::DW_LANG_C_plus_plus_11 ||
Language == dwarf::DW_LANG_C_plus_plus_14 ||
Language == dwarf::DW_LANG_ObjC_plus_plus ||
Language == dwarf::DW_LANG_C))
return Gsym.insertString(ShortName, false);
if (ShortName.startswith("_Z") &&
(ShortName.contains(".isra.") || ShortName.contains(".part.")))
return Gsym.insertString(ShortName, false);
DWARFDie ParentDeclCtxDie = GetParentDeclContextDIE(Die);
if (ParentDeclCtxDie) {
std::string Name = ShortName.str();
while (ParentDeclCtxDie) {
StringRef ParentName(ParentDeclCtxDie.getName(DINameKind::ShortName));
if (!ParentName.empty()) {
if (ParentName.front() == '<' && ParentName.back() == '>')
Name = "{" + ParentName.substr(1, ParentName.size() - 2).str() + "}" +
"::" + Name;
else
Name = ParentName.str() + "::" + Name;
}
ParentDeclCtxDie = GetParentDeclContextDIE(ParentDeclCtxDie);
}
return Gsym.insertString(Name, true);
}
return Gsym.insertString(ShortName, false);
}
static bool hasInlineInfo(DWARFDie Die, uint32_t Depth) {
bool CheckChildren = true;
switch (Die.getTag()) {
case dwarf::DW_TAG_subprogram:
CheckChildren = Depth == 0;
break;
case dwarf::DW_TAG_inlined_subroutine:
return true;
default:
break;
}
if (!CheckChildren)
return false;
for (DWARFDie ChildDie : Die.children()) {
if (hasInlineInfo(ChildDie, Depth + 1))
return true;
}
return false;
}
static void parseInlineInfo(GsymCreator &Gsym, CUInfo &CUI, DWARFDie Die,
uint32_t Depth, FunctionInfo &FI,
InlineInfo &parent) {
if (!hasInlineInfo(Die, Depth))
return;
dwarf::Tag Tag = Die.getTag();
if (Tag == dwarf::DW_TAG_inlined_subroutine) {
InlineInfo II;
DWARFAddressRange FuncRange =
DWARFAddressRange(FI.startAddress(), FI.endAddress());
Expected<DWARFAddressRangesVector> RangesOrError = Die.getAddressRanges();
if (RangesOrError) {
for (const DWARFAddressRange &Range : RangesOrError.get()) {
if (FuncRange.LowPC <= Range.LowPC && Range.HighPC <= FuncRange.HighPC)
II.Ranges.insert(AddressRange(Range.LowPC, Range.HighPC));
}
}
if (II.Ranges.empty())
return;
if (auto NameIndex = getQualifiedNameIndex(Die, CUI.Language, Gsym))
II.Name = *NameIndex;
II.CallFile = CUI.DWARFToGSYMFileIndex(
Gsym, dwarf::toUnsigned(Die.find(dwarf::DW_AT_call_file), 0));
II.CallLine = dwarf::toUnsigned(Die.find(dwarf::DW_AT_call_line), 0);
for (DWARFDie ChildDie : Die.children())
parseInlineInfo(Gsym, CUI, ChildDie, Depth + 1, FI, II);
parent.Children.emplace_back(std::move(II));
return;
}
if (Tag == dwarf::DW_TAG_subprogram || Tag == dwarf::DW_TAG_lexical_block) {
for (DWARFDie ChildDie : Die.children())
parseInlineInfo(Gsym, CUI, ChildDie, Depth + 1, FI, parent);
}
}
static void convertFunctionLineTable(raw_ostream &Log, CUInfo &CUI,
DWARFDie Die, GsymCreator &Gsym,
FunctionInfo &FI) {
std::vector<uint32_t> RowVector;
const uint64_t StartAddress = FI.startAddress();
const uint64_t EndAddress = FI.endAddress();
const uint64_t RangeSize = EndAddress - StartAddress;
const object::SectionedAddress SecAddress{
StartAddress, object::SectionedAddress::UndefSection};
if (!CUI.LineTable->lookupAddressRange(SecAddress, RangeSize, RowVector)) {
std::string FilePath = Die.getDeclFile(
DILineInfoSpecifier::FileLineInfoKind::AbsoluteFilePath);
if (FilePath.empty())
return;
if (auto Line =
dwarf::toUnsigned(Die.findRecursively({dwarf::DW_AT_decl_line}))) {
LineEntry LE(StartAddress, Gsym.insertFile(FilePath), *Line);
FI.OptLineTable = LineTable();
FI.OptLineTable->push(LE);
}
return;
}
FI.OptLineTable = LineTable();
DWARFDebugLine::Row PrevRow;
for (uint32_t RowIndex : RowVector) {
const DWARFDebugLine::Row &Row = CUI.LineTable->Rows[RowIndex];
const uint32_t FileIdx = CUI.DWARFToGSYMFileIndex(Gsym, Row.File);
uint64_t RowAddress = Row.Address.Address;
if (!FI.Range.contains(RowAddress)) {
if (RowAddress < FI.Range.start()) {
Log << "error: DIE has a start address whose LowPC is between the "
"line table Row[" << RowIndex << "] with address "
<< HEX64(RowAddress) << " and the next one.\n";
Die.dump(Log, 0, DIDumpOptions::getForSingleDIE());
RowAddress = FI.Range.start();
} else {
continue;
}
}
LineEntry LE(RowAddress, FileIdx, Row.Line);
if (RowIndex != RowVector[0] && Row.Address < PrevRow.Address) {
auto FirstLE = FI.OptLineTable->first();
if (FirstLE && *FirstLE == LE) {
if (!Gsym.isQuiet()) {
Log << "warning: duplicate line table detected for DIE:\n";
Die.dump(Log, 0, DIDumpOptions::getForSingleDIE());
}
} else {
Log << "error: line table has addresses that do not "
<< "monotonically increase:\n";
for (uint32_t RowIndex2 : RowVector) {
CUI.LineTable->Rows[RowIndex2].dump(Log);
}
Die.dump(Log, 0, DIDumpOptions::getForSingleDIE());
}
break;
}
auto LastLE = FI.OptLineTable->last();
if (LastLE && LastLE->File == FileIdx && LastLE->Line == Row.Line)
continue;
if (Row.EndSequence) {
PrevRow = DWARFDebugLine::Row();
} else {
FI.OptLineTable->push(LE);
PrevRow = Row;
}
}
if (FI.OptLineTable->empty())
FI.OptLineTable = llvm::None;
}
void DwarfTransformer::handleDie(raw_ostream &OS, CUInfo &CUI, DWARFDie Die) {
switch (Die.getTag()) {
case dwarf::DW_TAG_subprogram: {
Expected<DWARFAddressRangesVector> RangesOrError = Die.getAddressRanges();
if (!RangesOrError) {
consumeError(RangesOrError.takeError());
break;
}
const DWARFAddressRangesVector &Ranges = RangesOrError.get();
if (Ranges.empty())
break;
auto NameIndex = getQualifiedNameIndex(Die, CUI.Language, Gsym);
if (!NameIndex) {
OS << "error: function at " << HEX64(Die.getOffset())
<< " has no name\n ";
Die.dump(OS, 0, DIDumpOptions::getForSingleDIE());
break;
}
for (const DWARFAddressRange &Range : Ranges) {
if (Range.LowPC >= Range.HighPC || CUI.isHighestAddress(Range.LowPC))
break;
if (!Gsym.IsValidTextAddress(Range.LowPC)) {
if (Range.LowPC != 0) {
if (!Gsym.isQuiet()) {
OS << "warning: DIE has an address range whose start address is "
"not in any executable sections ("
<< *Gsym.GetValidTextRanges()
<< ") and will not be processed:\n";
Die.dump(OS, 0, DIDumpOptions::getForSingleDIE());
}
}
break;
}
FunctionInfo FI;
FI.Range = {Range.LowPC, Range.HighPC};
FI.Name = *NameIndex;
if (CUI.LineTable) {
convertFunctionLineTable(OS, CUI, Die, Gsym, FI);
}
if (hasInlineInfo(Die, 0)) {
FI.Inline = InlineInfo();
FI.Inline->Name = *NameIndex;
FI.Inline->Ranges.insert(FI.Range);
parseInlineInfo(Gsym, CUI, Die, 0, FI, *FI.Inline);
}
Gsym.addFunctionInfo(std::move(FI));
}
} break;
default:
break;
}
for (DWARFDie ChildDie : Die.children())
handleDie(OS, CUI, ChildDie);
}
Error DwarfTransformer::convert(uint32_t NumThreads) {
size_t NumBefore = Gsym.getNumFunctionInfos();
auto getDie = [&](DWARFUnit &DwarfUnit) -> DWARFDie {
DWARFDie ReturnDie = DwarfUnit.getUnitDIE(false);
if (llvm::Optional<uint64_t> DWOId = DwarfUnit.getDWOId()) {
DWARFUnit *DWOCU = DwarfUnit.getNonSkeletonUnitDIE(false).getDwarfUnit();
if (!DWOCU->isDWOUnit()) {
std::string DWOName = dwarf::toString(
DwarfUnit.getUnitDIE().find(
{dwarf::DW_AT_dwo_name, dwarf::DW_AT_GNU_dwo_name}),
"");
Log << "warning: Unable to retrieve DWO .debug_info section for "
<< DWOName << "\n";
} else {
ReturnDie = DWOCU->getUnitDIE(false);
}
}
return ReturnDie;
};
if (NumThreads == 1) {
for (const auto &CU : DICtx.compile_units()) {
DWARFDie Die = getDie(*CU);
CUInfo CUI(DICtx, dyn_cast<DWARFCompileUnit>(CU.get()));
handleDie(Log, CUI, Die);
}
} else {
for (const auto &CU : DICtx.compile_units())
CU->getAbbreviations();
ThreadPool pool(hardware_concurrency(NumThreads));
for (const auto &CU : DICtx.compile_units())
pool.async([&CU]() { CU->getUnitDIE(false ); });
pool.wait();
std::mutex LogMutex;
for (const auto &CU : DICtx.compile_units()) {
DWARFDie Die = getDie(*CU);
if (Die) {
CUInfo CUI(DICtx, dyn_cast<DWARFCompileUnit>(CU.get()));
pool.async([this, CUI, &LogMutex, Die]() mutable {
std::string ThreadLogStorage;
raw_string_ostream ThreadOS(ThreadLogStorage);
handleDie(ThreadOS, CUI, Die);
ThreadOS.flush();
if (!ThreadLogStorage.empty()) {
std::lock_guard<std::mutex> guard(LogMutex);
Log << ThreadLogStorage;
}
});
}
}
pool.wait();
}
size_t FunctionsAddedCount = Gsym.getNumFunctionInfos() - NumBefore;
Log << "Loaded " << FunctionsAddedCount << " functions from DWARF.\n";
return Error::success();
}
llvm::Error DwarfTransformer::verify(StringRef GsymPath) {
Log << "Verifying GSYM file \"" << GsymPath << "\":\n";
auto Gsym = GsymReader::openFile(GsymPath);
if (!Gsym)
return Gsym.takeError();
auto NumAddrs = Gsym->getNumAddresses();
DILineInfoSpecifier DLIS(
DILineInfoSpecifier::FileLineInfoKind::AbsoluteFilePath,
DILineInfoSpecifier::FunctionNameKind::LinkageName);
std::string gsymFilename;
for (uint32_t I = 0; I < NumAddrs; ++I) {
auto FuncAddr = Gsym->getAddress(I);
if (!FuncAddr)
return createStringError(std::errc::invalid_argument,
"failed to extract address[%i]", I);
auto FI = Gsym->getFunctionInfo(*FuncAddr);
if (!FI)
return createStringError(std::errc::invalid_argument,
"failed to extract function info for address 0x%"
PRIu64, *FuncAddr);
for (auto Addr = *FuncAddr; Addr < *FuncAddr + FI->size(); ++Addr) {
const object::SectionedAddress SectAddr{
Addr, object::SectionedAddress::UndefSection};
auto LR = Gsym->lookup(Addr);
if (!LR)
return LR.takeError();
auto DwarfInlineInfos =
DICtx.getInliningInfoForAddress(SectAddr, DLIS);
uint32_t NumDwarfInlineInfos = DwarfInlineInfos.getNumberOfFrames();
if (NumDwarfInlineInfos == 0) {
DwarfInlineInfos.addFrame(
DICtx.getLineInfoForAddress(SectAddr, DLIS));
}
if (NumDwarfInlineInfos == 1 &&
DwarfInlineInfos.getFrame(0).FileName == "<invalid>") {
DwarfInlineInfos = DIInliningInfo();
NumDwarfInlineInfos = 0;
}
if (NumDwarfInlineInfos > 0 &&
NumDwarfInlineInfos != LR->Locations.size()) {
Log << "error: address " << HEX64(Addr) << " has "
<< NumDwarfInlineInfos << " DWARF inline frames and GSYM has "
<< LR->Locations.size() << "\n";
Log << " " << NumDwarfInlineInfos << " DWARF frames:\n";
for (size_t Idx = 0; Idx < NumDwarfInlineInfos; ++Idx) {
const auto &dii = DwarfInlineInfos.getFrame(Idx);
Log << " [" << Idx << "]: " << dii.FunctionName << " @ "
<< dii.FileName << ':' << dii.Line << '\n';
}
Log << " " << LR->Locations.size() << " GSYM frames:\n";
for (size_t Idx = 0, count = LR->Locations.size();
Idx < count; ++Idx) {
const auto &gii = LR->Locations[Idx];
Log << " [" << Idx << "]: " << gii.Name << " @ " << gii.Dir
<< '/' << gii.Base << ':' << gii.Line << '\n';
}
DwarfInlineInfos = DICtx.getInliningInfoForAddress(SectAddr, DLIS);
Gsym->dump(Log, *FI);
continue;
}
for (size_t Idx = 0, count = LR->Locations.size(); Idx < count;
++Idx) {
const auto &gii = LR->Locations[Idx];
if (Idx < NumDwarfInlineInfos) {
const auto &dii = DwarfInlineInfos.getFrame(Idx);
gsymFilename = LR->getSourceFile(Idx);
if (dii.FunctionName.find(gii.Name.str()) != 0)
Log << "error: address " << HEX64(Addr) << " DWARF function \""
<< dii.FunctionName.c_str()
<< "\" doesn't match GSYM function \"" << gii.Name << "\"\n";
if (dii.FileName != gsymFilename)
Log << "error: address " << HEX64(Addr) << " DWARF path \""
<< dii.FileName.c_str() << "\" doesn't match GSYM path \""
<< gsymFilename.c_str() << "\"\n";
if (dii.Line != gii.Line)
Log << "error: address " << HEX64(Addr) << " DWARF line "
<< dii.Line << " != GSYM line " << gii.Line << "\n";
}
}
}
}
return Error::success();
}