[lld] 2a3a79c - [lld-macho][NFC] Preserve original symbol isec, unwindEntry and size (#88357)
via llvm-commits
llvm-commits at lists.llvm.org
Thu Apr 18 11:42:26 PDT 2024
Author: alx32
Date: 2024-04-18T11:42:22-07:00
New Revision: 2a3a79ce4c2149d7787d56f9841b66cacc9061d0
URL: https://github.com/llvm/llvm-project/commit/2a3a79ce4c2149d7787d56f9841b66cacc9061d0
DIFF: https://github.com/llvm/llvm-project/commit/2a3a79ce4c2149d7787d56f9841b66cacc9061d0.diff
LOG: [lld-macho][NFC] Preserve original symbol isec, unwindEntry and size (#88357)
Currently, when moving symbols from one `InputSection` to another (like
in ICF) we directly update the symbol's `isec`, `unwindEntry` and
`size`. By doing this we lose the original information. This information
will be needed in a future change. Since when moving symbols we always
set the symbol's `wasCoalesced` and `isec-> replacement`, we can just
use this info to conditionally get the information we need at access
time.
Added:
Modified:
lld/MachO/ICF.cpp
lld/MachO/InputFiles.cpp
lld/MachO/InputSection.cpp
lld/MachO/MapFile.cpp
lld/MachO/MarkLive.cpp
lld/MachO/ObjC.cpp
lld/MachO/Relocations.cpp
lld/MachO/SectionPriorities.cpp
lld/MachO/SymbolTable.cpp
lld/MachO/Symbols.cpp
lld/MachO/Symbols.h
lld/MachO/SyntheticSections.cpp
lld/MachO/UnwindInfoSection.cpp
lld/MachO/Writer.cpp
Removed:
################################################################################
diff --git a/lld/MachO/ICF.cpp b/lld/MachO/ICF.cpp
index 0278bf7c6751a2..fc786b571dc64f 100644
--- a/lld/MachO/ICF.cpp
+++ b/lld/MachO/ICF.cpp
@@ -133,13 +133,13 @@ bool ICF::equalsConstant(const ConcatInputSection *ia,
assert(isa<Defined>(sa));
const auto *da = cast<Defined>(sa);
const auto *db = cast<Defined>(sb);
- if (!da->isec || !db->isec) {
+ if (!da->isec() || !db->isec()) {
assert(da->isAbsolute() && db->isAbsolute());
return da->value + ra.addend == db->value + rb.addend;
}
- isecA = da->isec;
+ isecA = da->isec();
valueA = da->value;
- isecB = db->isec;
+ isecB = db->isec();
valueB = db->value;
} else {
isecA = ra.referent.get<InputSection *>();
@@ -191,10 +191,10 @@ bool ICF::equalsVariable(const ConcatInputSection *ia,
const auto *db = cast<Defined>(rb.referent.get<Symbol *>());
if (da->isAbsolute())
return true;
- isecA = dyn_cast<ConcatInputSection>(da->isec);
+ isecA = dyn_cast<ConcatInputSection>(da->isec());
if (!isecA)
return true; // literal sections were checked in equalsConstant.
- isecB = cast<ConcatInputSection>(db->isec);
+ isecB = cast<ConcatInputSection>(db->isec());
} else {
const auto *sa = ra.referent.get<InputSection *>();
const auto *sb = rb.referent.get<InputSection *>();
@@ -212,7 +212,7 @@ bool ICF::equalsVariable(const ConcatInputSection *ia,
// info matches. For simplicity, we only handle the case where there are only
// symbols at offset zero within the section (which is typically the case with
// .subsections_via_symbols.)
- auto hasUnwind = [](Defined *d) { return d->unwindEntry != nullptr; };
+ auto hasUnwind = [](Defined *d) { return d->unwindEntry() != nullptr; };
const auto *itA = llvm::find_if(ia->symbols, hasUnwind);
const auto *itB = llvm::find_if(ib->symbols, hasUnwind);
if (itA == ia->symbols.end())
@@ -221,8 +221,8 @@ bool ICF::equalsVariable(const ConcatInputSection *ia,
return false;
const Defined *da = *itA;
const Defined *db = *itB;
- if (da->unwindEntry->icfEqClass[icfPass % 2] !=
- db->unwindEntry->icfEqClass[icfPass % 2] ||
+ if (da->unwindEntry()->icfEqClass[icfPass % 2] !=
+ db->unwindEntry()->icfEqClass[icfPass % 2] ||
da->value != 0 || db->value != 0)
return false;
auto isZero = [](Defined *d) { return d->value == 0; };
@@ -289,13 +289,13 @@ void ICF::run() {
for (const Reloc &r : isec->relocs) {
if (auto *sym = r.referent.dyn_cast<Symbol *>()) {
if (auto *defined = dyn_cast<Defined>(sym)) {
- if (defined->isec) {
+ if (defined->isec()) {
if (auto *referentIsec =
- dyn_cast<ConcatInputSection>(defined->isec))
+ dyn_cast<ConcatInputSection>(defined->isec()))
hash += defined->value + referentIsec->icfEqClass[icfPass % 2];
else
- hash += defined->isec->kind() +
- defined->isec->getOffset(defined->value);
+ hash += defined->isec()->kind() +
+ defined->isec()->getOffset(defined->value);
} else {
hash += defined->value;
}
@@ -368,8 +368,8 @@ void ICF::segregate(size_t begin, size_t end, EqualsFn equals) {
void macho::markSymAsAddrSig(Symbol *s) {
if (auto *d = dyn_cast_or_null<Defined>(s))
- if (d->isec)
- d->isec->keepUnique = true;
+ if (d->isec())
+ d->isec()->keepUnique = true;
}
void macho::markAddrSigSymbols() {
@@ -430,8 +430,8 @@ void macho::foldIdenticalSections(bool onlyCfStrings) {
if (isFoldable) {
foldable.push_back(isec);
for (Defined *d : isec->symbols)
- if (d->unwindEntry)
- foldable.push_back(d->unwindEntry);
+ if (d->unwindEntry())
+ foldable.push_back(d->unwindEntry());
// Some sections have embedded addends that foil ICF's hashing / equality
// checks. (We can ignore embedded addends when doing ICF because the same
diff --git a/lld/MachO/InputFiles.cpp b/lld/MachO/InputFiles.cpp
index b36d390cc16ade..8d66b37534f47e 100644
--- a/lld/MachO/InputFiles.cpp
+++ b/lld/MachO/InputFiles.cpp
@@ -1170,7 +1170,7 @@ void ObjFile::registerCompactUnwind(Section &compactUnwindSection) {
continue;
}
add += sym->value;
- referentIsec = cast<ConcatInputSection>(sym->isec);
+ referentIsec = cast<ConcatInputSection>(sym->isec());
} else {
referentIsec =
cast<ConcatInputSection>(r.referent.dyn_cast<InputSection *>());
@@ -1191,7 +1191,7 @@ void ObjFile::registerCompactUnwind(Section &compactUnwindSection) {
++it;
continue;
}
- d->unwindEntry = isec;
+ d->originalUnwindEntry = isec;
// Now that the symbol points to the unwind entry, we can remove the reloc
// that points from the unwind entry back to the symbol.
//
@@ -1348,7 +1348,7 @@ targetSymFromCanonicalSubtractor(const InputSection *isec,
}
if (Invert)
std::swap(pcSym, target);
- if (pcSym->isec == isec) {
+ if (pcSym->isec() == isec) {
if (pcSym->value - (Invert ? -1 : 1) * minuend.addend != subtrahend.offset)
fatal("invalid FDE relocation in __eh_frame");
} else {
@@ -1420,7 +1420,7 @@ void ObjFile::registerEhFrames(Section &ehFrameSection) {
// We already have an explicit relocation for the CIE offset.
cieIsec =
targetSymFromCanonicalSubtractor</*Invert=*/true>(isec, cieOffRelocIt)
- ->isec;
+ ->isec();
dataOff += sizeof(uint32_t);
} else {
// If we haven't found a relocation, then the CIE offset is most likely
@@ -1480,15 +1480,15 @@ void ObjFile::registerEhFrames(Section &ehFrameSection) {
// to register the unwind entry under same symbol.
// This is not particularly efficient, but we should run into this case
// infrequently (only when handling the output of `ld -r`).
- if (funcSym->isec)
- funcSym = findSymbolAtOffset(cast<ConcatInputSection>(funcSym->isec),
+ if (funcSym->isec())
+ funcSym = findSymbolAtOffset(cast<ConcatInputSection>(funcSym->isec()),
funcSym->value);
} else {
funcSym = findSymbolAtAddress(sections, funcAddr);
ehRelocator.makePcRel(funcAddrOff, funcSym, target->p2WordSize);
}
// The symbol has been coalesced, or already has a compact unwind entry.
- if (!funcSym || funcSym->getFile() != this || funcSym->unwindEntry) {
+ if (!funcSym || funcSym->getFile() != this || funcSym->unwindEntry()) {
// We must prune unused FDEs for correctness, so we cannot rely on
// -dead_strip being enabled.
isec->live = false;
@@ -1497,7 +1497,8 @@ void ObjFile::registerEhFrames(Section &ehFrameSection) {
InputSection *lsdaIsec = nullptr;
if (lsdaAddrRelocIt != isec->relocs.end()) {
- lsdaIsec = targetSymFromCanonicalSubtractor(isec, lsdaAddrRelocIt)->isec;
+ lsdaIsec =
+ targetSymFromCanonicalSubtractor(isec, lsdaAddrRelocIt)->isec();
} else if (lsdaAddrOpt) {
uint64_t lsdaAddr = *lsdaAddrOpt;
Section *sec = findContainingSection(sections, &lsdaAddr);
@@ -1507,7 +1508,7 @@ void ObjFile::registerEhFrames(Section &ehFrameSection) {
}
fdes[isec] = {funcLength, cie.personalitySymbol, lsdaIsec};
- funcSym->unwindEntry = isec;
+ funcSym->originalUnwindEntry = isec;
ehRelocator.commit();
}
diff --git a/lld/MachO/InputSection.cpp b/lld/MachO/InputSection.cpp
index 5c1e07cd21b1fb..904701731684b3 100644
--- a/lld/MachO/InputSection.cpp
+++ b/lld/MachO/InputSection.cpp
@@ -194,10 +194,8 @@ void ConcatInputSection::foldIdentical(ConcatInputSection *copy) {
copy->live = false;
copy->wasCoalesced = true;
copy->replacement = this;
- for (auto ©Sym : copy->symbols) {
+ for (auto ©Sym : copy->symbols)
copySym->wasIdenticalCodeFolded = true;
- copySym->size = 0;
- }
symbols.insert(symbols.end(), copy->symbols.begin(), copy->symbols.end());
copy->symbols.clear();
@@ -207,7 +205,7 @@ void ConcatInputSection::foldIdentical(ConcatInputSection *copy) {
return;
for (auto it = symbols.begin() + 1; it != symbols.end(); ++it) {
assert((*it)->value == 0);
- (*it)->unwindEntry = nullptr;
+ (*it)->originalUnwindEntry = nullptr;
}
}
diff --git a/lld/MachO/MapFile.cpp b/lld/MachO/MapFile.cpp
index 2a31a5c09cdd22..5bcaeca48da2a2 100644
--- a/lld/MachO/MapFile.cpp
+++ b/lld/MachO/MapFile.cpp
@@ -77,8 +77,8 @@ static MapInfo gatherMapInfo() {
// Only emit the prevailing definition of a symbol. Also, don't emit
// the symbol if it is part of a cstring section (we use the literal
// value instead, similar to ld64)
- if (d->isec && d->getFile() == file &&
- !isa<CStringInputSection>(d->isec)) {
+ if (d->isec() && d->getFile() == file &&
+ !isa<CStringInputSection>(d->isec())) {
isReferencedFile = true;
if (!d->isLive())
info.deadSymbols.push_back(d);
@@ -155,6 +155,12 @@ static void printNonLazyPointerSection(raw_fd_ostream &os,
target->wordSize, sym->getName().str().data());
}
+static uint64_t getSymSizeForMap(Defined *sym) {
+ if (sym->wasIdenticalCodeFolded)
+ return 0;
+ return sym->size;
+}
+
void macho::writeMapFile() {
if (config->mapFile.empty())
return;
@@ -201,9 +207,10 @@ void macho::writeMapFile() {
auto printIsecArrSyms = [&](const std::vector<ConcatInputSection *> &arr) {
for (const ConcatInputSection *isec : arr) {
for (Defined *sym : isec->symbols) {
- if (!(isPrivateLabel(sym->getName()) && sym->size == 0))
+ if (!(isPrivateLabel(sym->getName()) && getSymSizeForMap(sym) == 0))
os << format("0x%08llX\t0x%08llX\t[%3u] %s\n", sym->getVA(),
- sym->size, readerToFileOrdinal[sym->getFile()],
+ getSymSizeForMap(sym),
+ readerToFileOrdinal[sym->getFile()],
sym->getName().str().data());
}
}
@@ -255,7 +262,7 @@ void macho::writeMapFile() {
os << "# \tSize \tFile Name\n";
for (Defined *sym : info.deadSymbols) {
assert(!sym->isLive());
- os << format("<<dead>>\t0x%08llX\t[%3u] %s\n", sym->size,
+ os << format("<<dead>>\t0x%08llX\t[%3u] %s\n", getSymSizeForMap(sym),
readerToFileOrdinal[sym->getFile()],
sym->getName().str().data());
}
diff --git a/lld/MachO/MarkLive.cpp b/lld/MachO/MarkLive.cpp
index a37213d5613afb..c26c3aa321197e 100644
--- a/lld/MachO/MarkLive.cpp
+++ b/lld/MachO/MarkLive.cpp
@@ -110,10 +110,10 @@ void MarkLiveImpl<RecordWhyLive>::addSym(
if (!config->whyLive.empty() && config->whyLive.match(s->getName()))
printWhyLive(s, prev);
if (auto *d = dyn_cast<Defined>(s)) {
- if (d->isec)
- enqueue(d->isec, d->value, prev);
- if (d->unwindEntry)
- enqueue(d->unwindEntry, 0, prev);
+ if (d->isec())
+ enqueue(d->isec(), d->value, prev);
+ if (d->unwindEntry())
+ enqueue(d->unwindEntry(), 0, prev);
}
}
@@ -179,7 +179,7 @@ void MarkLiveImpl<RecordWhyLive>::markTransitively() {
if (s->isLive()) {
InputSection *referentIsec = nullptr;
if (auto *d = dyn_cast<Defined>(s))
- referentIsec = d->isec;
+ referentIsec = d->isec();
enqueue(isec, 0, makeEntry(referentIsec, nullptr));
}
} else {
diff --git a/lld/MachO/ObjC.cpp b/lld/MachO/ObjC.cpp
index c28f2b42a72ce8..95fe0c9374f150 100644
--- a/lld/MachO/ObjC.cpp
+++ b/lld/MachO/ObjC.cpp
@@ -191,8 +191,8 @@ static StringRef getReferentString(const Reloc &r) {
if (auto *isec = r.referent.dyn_cast<InputSection *>())
return cast<CStringInputSection>(isec)->getStringRefAtOffset(r.addend);
auto *sym = cast<Defined>(r.referent.get<Symbol *>());
- return cast<CStringInputSection>(sym->isec)->getStringRefAtOffset(sym->value +
- r.addend);
+ return cast<CStringInputSection>(sym->isec())
+ ->getStringRefAtOffset(sym->value + r.addend);
}
void ObjcCategoryChecker::parseMethods(const ConcatInputSection *methodsIsec,
@@ -306,7 +306,7 @@ void ObjcCategoryChecker::parseClass(const Defined *classSym) {
return nullptr;
};
- const auto *classIsec = cast<ConcatInputSection>(classSym->isec);
+ const auto *classIsec = cast<ConcatInputSection>(classSym->isec());
// Parse instance methods.
if (const auto *instanceMethodsIsec = getMethodsIsec(classIsec))
@@ -314,7 +314,7 @@ void ObjcCategoryChecker::parseClass(const Defined *classSym) {
MK_Instance);
// Class methods are contained in the metaclass.
- if (const auto *r = classSym->isec->getRelocAt(classLayout.metaClassOffset))
+ if (const auto *r = classSym->isec()->getRelocAt(classLayout.metaClassOffset))
if (const auto *classMethodsIsec = getMethodsIsec(
cast<ConcatInputSection>(r->getReferentInputSection())))
parseMethods(classMethodsIsec, classSym, classIsec, MCK_Class, MK_Static);
@@ -561,9 +561,9 @@ void ObjcCategoryMerger::tryEraseDefinedAtIsecOffset(
if (!sym)
return;
- if (auto *cisec = dyn_cast_or_null<ConcatInputSection>(sym->isec))
+ if (auto *cisec = dyn_cast_or_null<ConcatInputSection>(sym->isec()))
eraseISec(cisec);
- else if (auto *csisec = dyn_cast_or_null<CStringInputSection>(sym->isec)) {
+ else if (auto *csisec = dyn_cast_or_null<CStringInputSection>(sym->isec())) {
uint32_t totalOffset = sym->value + reloc->addend;
StringPiece &piece = csisec->getStringPiece(totalOffset);
piece.live = false;
@@ -588,7 +588,7 @@ void ObjcCategoryMerger::collectCategoryWriterInfoFromCategory(
assert(catNameSym && "Category does not have a valid name Symbol");
collectSectionWriteInfoFromIsec<CStringSection>(
- catNameSym->isec, infoCategoryWriter.catNameInfo);
+ catNameSym->isec(), infoCategoryWriter.catNameInfo);
}
// Collect writer info from all the category lists (we're assuming they all
@@ -599,7 +599,7 @@ void ObjcCategoryMerger::collectCategoryWriterInfoFromCategory(
if (Defined *ptrList =
tryGetDefinedAtIsecOffset(catInfo.catBodyIsec, off)) {
collectSectionWriteInfoFromIsec<ConcatOutputSection>(
- ptrList->isec, infoCategoryWriter.catPtrListInfo);
+ ptrList->isec(), infoCategoryWriter.catPtrListInfo);
// we've successfully collected data, so we can break
break;
}
@@ -627,7 +627,7 @@ void ObjcCategoryMerger::parseProtocolListInfo(const ConcatInputSection *isec,
// platform pointer size, but to simplify implementation we always just read
// the lower 32b which should be good enough.
uint32_t protocolCount = *reinterpret_cast<const uint32_t *>(
- ptrListSym->isec->data.data() + listHeaderLayout.structSizeOffset);
+ ptrListSym->isec()->data.data() + listHeaderLayout.structSizeOffset);
ptrList.structCount += protocolCount;
ptrList.structSize = target->wordSize;
@@ -636,7 +636,7 @@ void ObjcCategoryMerger::parseProtocolListInfo(const ConcatInputSection *isec,
(protocolCount * target->wordSize) +
/*header(count)*/ protocolListHeaderLayout.totalSize +
/*extra null value*/ target->wordSize;
- assert(expectedListSize == ptrListSym->isec->data.size() &&
+ assert(expectedListSize == ptrListSym->isec()->data.size() &&
"Protocol list does not match expected size");
// Suppress unsuded var warning
@@ -644,7 +644,7 @@ void ObjcCategoryMerger::parseProtocolListInfo(const ConcatInputSection *isec,
uint32_t off = protocolListHeaderLayout.totalSize;
for (uint32_t inx = 0; inx < protocolCount; ++inx) {
- const Reloc *reloc = ptrListSym->isec->getRelocAt(off);
+ const Reloc *reloc = ptrListSym->isec()->getRelocAt(off);
assert(reloc && "No reloc found at protocol list offset");
auto *listSym = dyn_cast_or_null<Defined>(reloc->referent.get<Symbol *>());
@@ -653,7 +653,7 @@ void ObjcCategoryMerger::parseProtocolListInfo(const ConcatInputSection *isec,
ptrList.allPtrs.push_back(listSym);
off += target->wordSize;
}
- assert((ptrListSym->isec->getRelocAt(off) == nullptr) &&
+ assert((ptrListSym->isec()->getRelocAt(off) == nullptr) &&
"expected null terminating protocol");
assert(off + /*extra null value*/ target->wordSize == expectedListSize &&
"Protocol list end offset does not match expected size");
@@ -678,9 +678,9 @@ void ObjcCategoryMerger::parsePointerListInfo(const ConcatInputSection *isec,
assert(ptrListSym && "Reloc does not have a valid Defined");
uint32_t thisStructSize = *reinterpret_cast<const uint32_t *>(
- ptrListSym->isec->data.data() + listHeaderLayout.structSizeOffset);
+ ptrListSym->isec()->data.data() + listHeaderLayout.structSizeOffset);
uint32_t thisStructCount = *reinterpret_cast<const uint32_t *>(
- ptrListSym->isec->data.data() + listHeaderLayout.structCountOffset);
+ ptrListSym->isec()->data.data() + listHeaderLayout.structCountOffset);
assert(thisStructSize == ptrList.pointersPerStruct * target->wordSize);
assert(!ptrList.structSize || (thisStructSize == ptrList.structSize));
@@ -690,12 +690,12 @@ void ObjcCategoryMerger::parsePointerListInfo(const ConcatInputSection *isec,
uint32_t expectedListSize =
listHeaderLayout.totalSize + (thisStructSize * thisStructCount);
- assert(expectedListSize == ptrListSym->isec->data.size() &&
+ assert(expectedListSize == ptrListSym->isec()->data.size() &&
"Pointer list does not match expected size");
for (uint32_t off = listHeaderLayout.totalSize; off < expectedListSize;
off += target->wordSize) {
- const Reloc *reloc = ptrListSym->isec->getRelocAt(off);
+ const Reloc *reloc = ptrListSym->isec()->getRelocAt(off);
assert(reloc && "No reloc found at pointer list offset");
auto *listSym = dyn_cast_or_null<Defined>(reloc->referent.get<Symbol *>());
@@ -1054,7 +1054,7 @@ void ObjcCategoryMerger::createSymbolReference(Defined *refFrom,
r.offset = offset;
r.addend = 0;
r.referent = const_cast<Symbol *>(refTo);
- refFrom->isec->relocs.push_back(r);
+ refFrom->isec()->relocs.push_back(r);
}
void ObjcCategoryMerger::collectAndValidateCategoriesData() {
@@ -1076,7 +1076,7 @@ void ObjcCategoryMerger::collectAndValidateCategoriesData() {
if (!categorySym->getName().starts_with(objc::symbol_names::category))
continue;
- auto *catBodyIsec = dyn_cast<ConcatInputSection>(categorySym->isec);
+ auto *catBodyIsec = dyn_cast<ConcatInputSection>(categorySym->isec());
assert(catBodyIsec &&
"Category data section is not an ConcatInputSection");
diff --git a/lld/MachO/Relocations.cpp b/lld/MachO/Relocations.cpp
index 4e840c6912cc57..afe7f454e6a230 100644
--- a/lld/MachO/Relocations.cpp
+++ b/lld/MachO/Relocations.cpp
@@ -24,7 +24,7 @@ static_assert(sizeof(void *) != 8 || sizeof(Reloc) == 24,
InputSection *Reloc::getReferentInputSection() const {
if (const auto *sym = referent.dyn_cast<Symbol *>()) {
if (const auto *d = dyn_cast<Defined>(sym))
- return d->isec;
+ return d->isec();
return nullptr;
} else {
return referent.get<InputSection *>();
diff --git a/lld/MachO/SectionPriorities.cpp b/lld/MachO/SectionPriorities.cpp
index 976ea03980e963..907aee29d2386f 100644
--- a/lld/MachO/SectionPriorities.cpp
+++ b/lld/MachO/SectionPriorities.cpp
@@ -236,7 +236,7 @@ DenseMap<const InputSection *, size_t> CallGraphSort::run() {
// section.
for (Symbol *sym : isec->getFile()->symbols) {
if (auto *d = dyn_cast_or_null<Defined>(sym)) {
- if (d->isec == isec)
+ if (d->isec() == isec)
os << sym->getName() << "\n";
}
}
@@ -258,7 +258,7 @@ macho::PriorityBuilder::getSymbolPriority(const Defined *sym) {
if (it == priorities.end())
return std::nullopt;
const SymbolPriorityEntry &entry = it->second;
- const InputFile *f = sym->isec->getFile();
+ const InputFile *f = sym->isec()->getFile();
if (!f)
return entry.anyObjectFile;
// We don't use toString(InputFile *) here because it returns the full path
@@ -287,7 +287,7 @@ void macho::PriorityBuilder::extractCallGraphProfile() {
if (fromSym && toSym &&
(!hasOrderFile ||
(!getSymbolPriority(fromSym) && !getSymbolPriority(toSym))))
- callGraphProfile[{fromSym->isec, toSym->isec}] += entry.count;
+ callGraphProfile[{fromSym->isec(), toSym->isec()}] += entry.count;
}
}
}
@@ -370,7 +370,7 @@ macho::PriorityBuilder::buildInputSectionPriorities() {
std::optional<size_t> symbolPriority = getSymbolPriority(sym);
if (!symbolPriority)
return;
- size_t &priority = sectionPriorities[sym->isec];
+ size_t &priority = sectionPriorities[sym->isec()];
priority = std::max(priority, *symbolPriority);
};
diff --git a/lld/MachO/SymbolTable.cpp b/lld/MachO/SymbolTable.cpp
index 755ff270e2f7a9..f0a92da8777e13 100644
--- a/lld/MachO/SymbolTable.cpp
+++ b/lld/MachO/SymbolTable.cpp
@@ -80,14 +80,14 @@ static void transplantSymbolsAtOffset(InputSection *fromIsec,
// iterator. However, that is typically the case for files that have
// .subsections_via_symbols set.
insertIt = toIsec->symbols.insert(insertIt, d);
- d->isec = toIsec;
+ d->originalIsec = toIsec;
d->value = toOff;
// We don't want to have more than one unwindEntry at a given address, so
// drop the redundant ones. We We can safely drop the unwindEntries of
// the symbols in fromIsec since we will be adding another unwindEntry as
// we finish parsing toIsec's file. (We can assume that toIsec has its
// own unwindEntry because of the ODR.)
- d->unwindEntry = nullptr;
+ d->originalUnwindEntry = nullptr;
}
return true;
});
@@ -121,8 +121,8 @@ Defined *SymbolTable::addDefined(StringRef name, InputFile *file,
// in ObjFile::parseSymbols() such that extern weak symbols appear
// last, so we don't need to worry about subsequent symbols being
// added to an already-coalesced section.
- if (defined->isec)
- transplantSymbolsAtOffset(concatIsec, defined->isec,
+ if (defined->isec())
+ transplantSymbolsAtOffset(concatIsec, defined->isec(),
/*skip=*/nullptr, value, defined->value);
}
return defined;
@@ -130,7 +130,7 @@ Defined *SymbolTable::addDefined(StringRef name, InputFile *file,
if (defined->isWeakDef()) {
if (auto concatIsec =
- dyn_cast_or_null<ConcatInputSection>(defined->isec)) {
+ dyn_cast_or_null<ConcatInputSection>(defined->isec())) {
concatIsec->wasCoalesced = true;
if (isec)
transplantSymbolsAtOffset(concatIsec, isec, defined, defined->value,
@@ -212,7 +212,7 @@ Defined *SymbolTable::addDefined(StringRef name, InputFile *file,
Defined *SymbolTable::aliasDefined(Defined *src, StringRef target,
InputFile *newFile, bool makePrivateExtern) {
bool isPrivateExtern = makePrivateExtern || src->privateExtern;
- return addDefined(target, newFile, src->isec, src->value, src->size,
+ return addDefined(target, newFile, src->isec(), src->value, src->size,
src->isWeakDef(), isPrivateExtern,
src->referencedDynamically, src->noDeadStrip,
src->weakDefCanBeHidden);
diff --git a/lld/MachO/Symbols.cpp b/lld/MachO/Symbols.cpp
index 4428bd4d9dc909..7bac78a59e4fec 100644
--- a/lld/MachO/Symbols.cpp
+++ b/lld/MachO/Symbols.cpp
@@ -63,8 +63,8 @@ Defined::Defined(StringRefZ name, InputFile *file, InputSection *isec,
wasIdenticalCodeFolded(false),
referencedDynamically(isReferencedDynamically), noDeadStrip(noDeadStrip),
interposable(interposable), weakDefCanBeHidden(isWeakDefCanBeHidden),
- weakDef(isWeakDef), external(isExternal), isec(isec), value(value),
- size(size) {
+ weakDef(isWeakDef), external(isExternal), originalIsec(isec),
+ value(value), size(size) {
if (isec) {
isec->symbols.push_back(this);
// Maintain sorted order.
@@ -82,7 +82,7 @@ Defined::Defined(StringRefZ name, InputFile *file, InputSection *isec,
}
bool Defined::isTlv() const {
- return !isAbsolute() && isThreadLocalVariables(isec->getFlags());
+ return !isAbsolute() && isThreadLocalVariables(originalIsec->getFlags());
}
uint64_t Defined::getVA() const {
@@ -91,7 +91,7 @@ uint64_t Defined::getVA() const {
if (isAbsolute())
return value;
- if (!isec->isFinal) {
+ if (!isec()->isFinal) {
// A target arch that does not use thunks ought never ask for
// the address of a function that has not yet been finalized.
assert(target->usesThunks());
@@ -102,24 +102,28 @@ uint64_t Defined::getVA() const {
// expedient to return a contrived out-of-range address.
return TargetInfo::outOfRangeVA;
}
- return isec->getVA(value);
+ return isec()->getVA(value);
}
ObjFile *Defined::getObjectFile() const {
- return isec ? dyn_cast_or_null<ObjFile>(isec->getFile()) : nullptr;
-}
-
-void Defined::canonicalize() {
- if (unwindEntry)
- unwindEntry = unwindEntry->canonical();
- if (isec)
- isec = isec->canonical();
+ return originalIsec ? dyn_cast_or_null<ObjFile>(originalIsec->getFile())
+ : nullptr;
}
std::string Defined::getSourceLocation() {
- if (!isec)
+ if (!originalIsec)
return {};
- return isec->getSourceLocation(value);
+ return originalIsec->getSourceLocation(value);
+}
+
+// Get the canonical InputSection of the symbol.
+InputSection *Defined::isec() const {
+ return originalIsec ? originalIsec->canonical() : nullptr;
+}
+
+// Get the canonical unwind entry of the symbol.
+ConcatInputSection *Defined::unwindEntry() const {
+ return originalUnwindEntry ? originalUnwindEntry->canonical() : nullptr;
}
uint64_t DylibSymbol::getVA() const {
diff --git a/lld/MachO/Symbols.h b/lld/MachO/Symbols.h
index f30294dbea9cb2..be5d135b8cd54f 100644
--- a/lld/MachO/Symbols.h
+++ b/lld/MachO/Symbols.h
@@ -129,7 +129,7 @@ class Defined : public Symbol {
bool isTlv() const override;
bool isExternal() const { return external; }
- bool isAbsolute() const { return isec == nullptr; }
+ bool isAbsolute() const { return originalIsec == nullptr; }
uint64_t getVA() const override;
@@ -139,9 +139,11 @@ class Defined : public Symbol {
std::string getSourceLocation();
- // Ensure this symbol's pointers to InputSections point to their canonical
- // copies.
- void canonicalize();
+ // Get the canonical InputSection of the symbol.
+ InputSection *isec() const;
+
+ // Get the canonical unwind entry of the symbol.
+ ConcatInputSection *unwindEntry() const;
static bool classof(const Symbol *s) { return s->kind() == DefinedKind; }
@@ -182,14 +184,17 @@ class Defined : public Symbol {
const bool external : 1;
public:
- InputSection *isec;
+ // The native InputSection of the symbol. The symbol may be moved to another
+ // InputSection in which case originalIsec->canonical() will point to the new
+ // InputSection
+ InputSection *originalIsec;
// Contains the offset from the containing subsection. Note that this is
//
diff erent from nlist::n_value, which is the absolute address of the symbol.
uint64_t value;
// size is only calculated for regular (non-bitcode) symbols.
uint64_t size;
// This can be a subsection of either __compact_unwind or __eh_frame.
- ConcatInputSection *unwindEntry = nullptr;
+ ConcatInputSection *originalUnwindEntry = nullptr;
};
// This enum does double-duty: as a symbol property, it indicates whether & how
diff --git a/lld/MachO/SyntheticSections.cpp b/lld/MachO/SyntheticSections.cpp
index 6f6b66118b7a94..29070810bb049e 100644
--- a/lld/MachO/SyntheticSections.cpp
+++ b/lld/MachO/SyntheticSections.cpp
@@ -828,7 +828,7 @@ void ObjCSelRefsHelper::initialize() {
auto Reloc = isec->relocs[0];
if (const auto *sym = Reloc.referent.dyn_cast<Symbol *>()) {
if (const auto *d = dyn_cast<Defined>(sym)) {
- auto *cisec = cast<CStringInputSection>(d->isec);
+ auto *cisec = cast<CStringInputSection>(d->isec());
auto methname = cisec->getStringRefAtOffset(d->value);
methnameToSelref[CachedHashStringRef(methname)] = isec;
}
@@ -1127,7 +1127,7 @@ void FunctionStartsSection::finalizeContents() {
if (auto *objFile = dyn_cast<ObjFile>(file)) {
for (const Symbol *sym : objFile->symbols) {
if (const auto *defined = dyn_cast_or_null<Defined>(sym)) {
- if (!defined->isec || !isCodeSection(defined->isec) ||
+ if (!defined->isec() || !isCodeSection(defined->isec()) ||
!defined->isLive())
continue;
addrs.push_back(defined->getVA());
@@ -1228,7 +1228,7 @@ void SymtabSection::emitStabs() {
if (!file || !file->compileUnit)
continue;
- symbolsNeedingStabs.emplace_back(defined, defined->isec->getFile()->id);
+ symbolsNeedingStabs.emplace_back(defined, defined->isec()->getFile()->id);
}
}
@@ -1243,7 +1243,7 @@ void SymtabSection::emitStabs() {
InputFile *lastFile = nullptr;
for (SortingPair &pair : symbolsNeedingStabs) {
Defined *defined = pair.first;
- InputSection *isec = defined->isec;
+ InputSection *isec = defined->isec();
ObjFile *file = cast<ObjFile>(isec->getFile());
if (lastFile == nullptr || lastFile != file) {
@@ -1256,7 +1256,7 @@ void SymtabSection::emitStabs() {
}
StabsEntry symStab;
- symStab.sect = defined->isec->parent->index;
+ symStab.sect = defined->isec()->parent->index;
symStab.strx = stringTableSection.addString(defined->getName());
symStab.value = defined->getVA();
@@ -1407,7 +1407,7 @@ template <class LP> void SymtabSectionImpl<LP>::writeTo(uint8_t *buf) const {
nList->n_value = defined->value;
} else {
nList->n_type = scope | N_SECT;
- nList->n_sect = defined->isec->parent->index;
+ nList->n_sect = defined->isec()->parent->index;
// For the N_SECT symbol type, n_value is the address of the symbol
nList->n_value = defined->getVA();
}
@@ -2000,7 +2000,7 @@ void ObjCMethListSection::setUp() {
assert(reloc && "Relocation expected at method list name slot");
auto *def = dyn_cast_or_null<Defined>(reloc->referent.get<Symbol *>());
assert(def && "Expected valid Defined at method list name slot");
- auto *cisec = cast<CStringInputSection>(def->isec);
+ auto *cisec = cast<CStringInputSection>(def->isec());
assert(cisec && "Expected method name to be in a CStringInputSection");
auto methname = cisec->getStringRefAtOffset(def->value);
if (!ObjCSelRefsHelper::getSelRef(methname))
@@ -2107,7 +2107,7 @@ void ObjCMethListSection::writeRelativeOffsetForIsec(
uint32_t symVA = def->getVA();
if (useSelRef) {
- auto *cisec = cast<CStringInputSection>(def->isec);
+ auto *cisec = cast<CStringInputSection>(def->isec());
auto methname = cisec->getStringRefAtOffset(def->value);
ConcatInputSection *selRef = ObjCSelRefsHelper::getSelRef(methname);
assert(selRef && "Expected all selector names to already be already be "
diff --git a/lld/MachO/UnwindInfoSection.cpp b/lld/MachO/UnwindInfoSection.cpp
index 1466e7b4fb394f..0ac2f39a6180c7 100644
--- a/lld/MachO/UnwindInfoSection.cpp
+++ b/lld/MachO/UnwindInfoSection.cpp
@@ -185,16 +185,16 @@ UnwindInfoSection::UnwindInfoSection()
// function symbols for each unique address regardless of whether they have
// associated unwind info.
void UnwindInfoSection::addSymbol(const Defined *d) {
- if (d->unwindEntry)
+ if (d->unwindEntry())
allEntriesAreOmitted = false;
// We don't yet know the final output address of this symbol, but we know that
// they are uniquely determined by a combination of the isec and value, so
// we use that as the key here.
- auto p = symbols.insert({{d->isec, d->value}, d});
+ auto p = symbols.insert({{d->isec(), d->value}, d});
// If we have multiple symbols at the same address, only one of them can have
// an associated unwind entry.
- if (!p.second && d->unwindEntry) {
- assert(p.first->second == d || !p.first->second->unwindEntry);
+ if (!p.second && d->unwindEntry()) {
+ assert(p.first->second == d || !p.first->second->unwindEntry());
p.first->second = d;
}
}
@@ -204,16 +204,16 @@ void UnwindInfoSectionImpl::prepare() {
// entries to the GOT. Hence the use of a MapVector for
// UnwindInfoSection::symbols.
for (const Defined *d : make_second_range(symbols))
- if (d->unwindEntry) {
- if (d->unwindEntry->getName() == section_names::compactUnwind) {
- prepareRelocations(d->unwindEntry);
+ if (d->unwindEntry()) {
+ if (d->unwindEntry()->getName() == section_names::compactUnwind) {
+ prepareRelocations(d->unwindEntry());
} else {
// We don't have to add entries to the GOT here because FDEs have
// explicit GOT relocations, so Writer::scanRelocations() will add those
// GOT entries. However, we still need to canonicalize the personality
// pointers (like prepareRelocations() does for CU entries) in order
// to avoid overflowing the 3-personality limit.
- FDE &fde = cast<ObjFile>(d->getFile())->fdes[d->unwindEntry];
+ FDE &fde = cast<ObjFile>(d->getFile())->fdes[d->unwindEntry()];
fde.personality = canonicalizePersonality(fde.personality);
}
}
@@ -279,7 +279,7 @@ void UnwindInfoSectionImpl::prepareRelocations(ConcatInputSection *isec) {
if (auto *defined = dyn_cast<Defined>(s)) {
// Check if we have created a synthetic symbol at the same address.
Symbol *&personality =
- personalityTable[{defined->isec, defined->value}];
+ personalityTable[{defined->isec(), defined->value}];
if (personality == nullptr) {
personality = defined;
in.got->addEntry(defined);
@@ -321,7 +321,7 @@ void UnwindInfoSectionImpl::prepareRelocations(ConcatInputSection *isec) {
Symbol *UnwindInfoSectionImpl::canonicalizePersonality(Symbol *personality) {
if (auto *defined = dyn_cast_or_null<Defined>(personality)) {
// Check if we have created a synthetic symbol at the same address.
- Symbol *&synth = personalityTable[{defined->isec, defined->value}];
+ Symbol *&synth = personalityTable[{defined->isec(), defined->value}];
if (synth == nullptr)
synth = defined;
else if (synth != defined)
@@ -340,12 +340,12 @@ void UnwindInfoSectionImpl::relocateCompactUnwind(
CompactUnwindEntry &cu = cuEntries[i];
const Defined *d = symbolsVec[i].second;
cu.functionAddress = d->getVA();
- if (!d->unwindEntry)
+ if (!d->unwindEntry())
return;
// If we have DWARF unwind info, create a slimmed-down CU entry that points
// to it.
- if (d->unwindEntry->getName() == section_names::ehFrame) {
+ if (d->unwindEntry()->getName() == section_names::ehFrame) {
// The unwinder will look for the DWARF entry starting at the hint,
// assuming the hint points to a valid CFI record start. If it
// fails to find the record, it proceeds in a linear search through the
@@ -355,11 +355,11 @@ void UnwindInfoSectionImpl::relocateCompactUnwind(
// but since we don't keep track of that, just encode zero -- the start of
// the section is always the start of a CFI record.
uint64_t dwarfOffsetHint =
- d->unwindEntry->outSecOff <= DWARF_SECTION_OFFSET
- ? d->unwindEntry->outSecOff
+ d->unwindEntry()->outSecOff <= DWARF_SECTION_OFFSET
+ ? d->unwindEntry()->outSecOff
: 0;
cu.encoding = target->modeDwarfEncoding | dwarfOffsetHint;
- const FDE &fde = cast<ObjFile>(d->getFile())->fdes[d->unwindEntry];
+ const FDE &fde = cast<ObjFile>(d->getFile())->fdes[d->unwindEntry()];
cu.functionLength = fde.funcLength;
// Omit the DWARF personality from compact-unwind entry so that we
// don't need to encode it.
@@ -368,14 +368,15 @@ void UnwindInfoSectionImpl::relocateCompactUnwind(
return;
}
- assert(d->unwindEntry->getName() == section_names::compactUnwind);
+ assert(d->unwindEntry()->getName() == section_names::compactUnwind);
- auto buf = reinterpret_cast<const uint8_t *>(d->unwindEntry->data.data()) -
- target->wordSize;
+ auto buf =
+ reinterpret_cast<const uint8_t *>(d->unwindEntry()->data.data()) -
+ target->wordSize;
cu.functionLength =
support::endian::read32le(buf + cuLayout.functionLengthOffset);
cu.encoding = support::endian::read32le(buf + cuLayout.encodingOffset);
- for (const Reloc &r : d->unwindEntry->relocs) {
+ for (const Reloc &r : d->unwindEntry()->relocs) {
if (r.offset == cuLayout.personalityOffset)
cu.personality = r.referent.get<Symbol *>();
else if (r.offset == cuLayout.lsdaOffset)
diff --git a/lld/MachO/Writer.cpp b/lld/MachO/Writer.cpp
index 1c054912551e3e..b9fcb45ef86b27 100644
--- a/lld/MachO/Writer.cpp
+++ b/lld/MachO/Writer.cpp
@@ -725,10 +725,9 @@ void Writer::scanSymbols() {
if (auto *defined = dyn_cast<Defined>(sym)) {
if (!defined->isLive())
continue;
- defined->canonicalize();
if (defined->overridesWeakDef)
addNonWeakDefinition(defined);
- if (!defined->isAbsolute() && isCodeSection(defined->isec))
+ if (!defined->isAbsolute() && isCodeSection(defined->isec()))
in.unwindInfo->addSymbol(defined);
} else if (const auto *dysym = dyn_cast<DylibSymbol>(sym)) {
// This branch intentionally doesn't check isLive().
@@ -756,9 +755,8 @@ void Writer::scanSymbols() {
if (auto *defined = dyn_cast_or_null<Defined>(sym)) {
if (!defined->isLive())
continue;
- defined->canonicalize();
if (!defined->isExternal() && !defined->isAbsolute() &&
- isCodeSection(defined->isec))
+ isCodeSection(defined->isec()))
in.unwindInfo->addSymbol(defined);
}
}
More information about the llvm-commits
mailing list