[ELF] Avoid referencing SectionBase::repl after ICF

[MaskRay]

Summary

It is fairly easy to forget SectionBase::repl after ICF.
Let ICF rewrite a Defined symbol's section field to avoid references to
SectionBase::repl in subsequent passes. This slightly improves the --icf=none
performance due to less indirection (maybe for --icf={safe,all} as well if most
symbols are Defined).

With this change, there is only one reference to repl (--gdb-index D89751).
We can undo f4fb5fd7523f8e3c3b3966d43c0a28457b59d1d8 (Move Repl to SectionBase.)
but move repl to InputSection instead.

[ikudrin]

LGTM

lld/ELF/InputSection.cpp

951–952

The comment has to be updated.

[MaskRay]

update comment

[MaskRay]

Move repl to InputSection

lld/ELF/CallGraphSort.cpp

if (res.second) {
       sections.push_back(isec);
       clusters.emplace_back(clusters.size(), isec->getSize());
     }
     return res.first->second;
   };
 
   // Create the graph.
   for (std::pair<SectionPair, uint64_t> &c : profile) {
-    const auto *fromSB = cast<InputSectionBase>(c.first.first->repl);
-    const auto *toSB = cast<InputSectionBase>(c.first.second->repl);
+    const auto *fromSB = cast<InputSectionBase>(c.first.first);
+    const auto *toSB = cast<InputSectionBase>(c.first.second);
     uint64_t weight = c.second;
 
     // Ignore edges between input sections belonging to different output
     // sections.  This is done because otherwise we would end up with clusters
     // containing input sections that can't actually be placed adjacently in the
     // output.  This messes with the cluster size and density calculations.  We
     // would also end up moving input sections in other output sections without
     // moving them closer to what calls them.

lld/ELF/ICF.cpp

for (size_t i = begin + 1; i < end; ++i) {
       // At this point we know sections merged are fully identical and hence
       // we want to remove duplicate implicit dependencies such as link order
       // and relocation sections.
       for (InputSection *isec : sections[i]->dependentSections)
         isec->markDead();
     }
   });
 
+  // Change Defined symbol's section field to the canonical one.
+  auto fold = [](Symbol *sym) {
+    if (auto *d = dyn_cast<Defined>(sym))
+      if (isa_and_nonnull<InputSectionBase>(d->section) &&
+          cast<InputSectionBase>(d->section)->repl != d->section) {
+        d->section = cast<InputSectionBase>(d->section)->repl;
+        d->folded = true;
+      }
+  };
+  for (Symbol *sym : symtab->symbols())
+    fold(sym);
+  for (InputFile *file : objectFiles)
+    for (Symbol *sym : cast<ELFFileBase>(file)->getLocalSymbols())
+      fold(sym);
+
   // InputSectionDescription::sections is populated by processSectionCommands().
   // ICF may fold some input sections assigned to output sections. Remove them.
   for (SectionCommand *cmd : script->sectionCommands)
     if (auto *sec = dyn_cast<OutputSection>(cmd))
       for (SectionCommand *subCmd : sec->commands)
         if (auto *isd = dyn_cast<InputSectionDescription>(subCmd))
           llvm::erase_if(isd->sections,
                          [](InputSection *isec) { return !isec->isLive(); });

lld/ELF/InputSection.h

 class SectionBase {
 public:
   enum Kind { Regular, EHFrame, Merge, Synthetic, Output };
 
   Kind kind() const { return (Kind)sectionKind; }
 
   StringRef name;
 
-  // This pointer points to the "real" instance of this instance.
-  // Usually Repl == this. However, if ICF merges two sections,
-  // Repl pointer of one section points to another section. So,
-  // if you need to get a pointer to this instance, do not use
-  // this but instead this->Repl.
-  SectionBase *repl;
-
   uint8_t sectionKind : 3;
 
   // The next two bit fields are only used by InputSectionBase, but we
   // put them here so the struct packs better.
 
   uint8_t bss : 1;
 
   // Set for sections that should not be folded by ICF.
public:
   bool isLive() const { return partition != 0; }
   void markLive() { partition = 1; }
   void markDead() { partition = 0; }
 
 protected:
   constexpr SectionBase(Kind sectionKind, StringRef name, uint64_t flags,
                         uint32_t entsize, uint32_t alignment, uint32_t type,
                         uint32_t info, uint32_t link)
-      : name(name), repl(this), sectionKind(sectionKind), bss(false),
-        keepUnique(false), partition(0), alignment(alignment), flags(flags),
-        entsize(entsize), type(type), link(link), info(info) {}
+      : name(name), sectionKind(sectionKind), bss(false), keepUnique(false),
+        partition(0), alignment(alignment), flags(flags), entsize(entsize),
+        type(type), link(link), info(info) {}
 };
 
 // This corresponds to a section of an input file.
 class InputSectionBase : public SectionBase {
 public:
   template <class ELFT>
   InputSectionBase(ObjFile<ELFT> &file, const typename ELFT::Shdr &header,
                    StringRef name, Kind sectionKind);
public:
   uint64_t getOffsetInFile() const;
 
   // Input sections are part of an output section. Special sections
   // like .eh_frame and merge sections are first combined into a
   // synthetic section that is then added to an output section. In all
   // cases this points one level up.
   SectionBase *parent = nullptr;
 
+  // This pointer points to the "real" instance of this instance.
+  // Usually repl == this. However, if ICF merges two sections,
+  // Repl pointer of one section points to another section. So,
+  // if you need to get a pointer to this instance, do not use
+  // this but instead this->Repl.
+  SectionBase *repl;
+
   // The next member in the section group if this section is in a group. This is
   // used by --gc-sections.
   InputSectionBase *nextInSectionGroup = nullptr;
 
   template <class ELFT> RelsOrRelas<ELFT> relsOrRelas() const;
 
   // InputSections that are dependent on us (reverse dependency for GC)
   llvm::TinyPtrVector<InputSection *> dependentSections;

lld/ELF/InputSection.cpp

 InputSectionBase::InputSectionBase(InputFile *file, uint64_t flags,
                                    uint32_t type, uint64_t entsize,
                                    uint32_t link, uint32_t info,
                                    uint32_t alignment, ArrayRef<uint8_t> data,
                                    StringRef name, Kind sectionKind)
     : SectionBase(sectionKind, name, flags, entsize, alignment, type, info,
                   link),
       file(file), rawData(data) {
+  repl = this;
   // In order to reduce memory allocation, we assume that mergeable
   // sections are smaller than 4 GiB, which is not an unreasonable
   // assumption as of 2017.
   if (sectionKind == SectionBase::Merge && rawData.size() > UINT32_MAX)
     error(toString(this) + ": section too large");
 
   // The ELF spec states that a value of 0 means the section has
   // no alignment constraints.
if (sym.type == STT_SECTION) {
               CHECK(file->getObj().sections(), file)[secIdx];
           warn("relocation refers to a discarded section: " +
                CHECK(file->getObj().getSectionName(sec), file) +
                "\n>>> referenced by " + getObjMsg(p->r_offset));
         }
         p->setSymbolAndType(0, 0, false);
         continue;
       }
-      SectionBase *section = d->section->repl;
+      SectionBase *section = d->section;
       if (!section->isLive()) {
         p->setSymbolAndType(0, 0, false);
         continue;
       }
 
       int64_t addend = getAddend<ELFT>(rel);
       const uint8_t *bufLoc = sec->data().begin() + rel.r_offset;
       if (!RelTy::IsRela)
if (tombstone ||
       // to resolve an address attribute (which may have a non-zero addend) to
       // -1+addend (wrap around to a low address).
       //
       // R_DTPREL type relocations represent an offset into the dynamic thread
       // vector. The computed value is st_value plus a non-negative offset.
       // Negative values are invalid, so -1 can be used as the tombstone value.
       //
       // If the referenced symbol is discarded (made Undefined), or the
       // section defining the referenced symbol is garbage collected,
-      // sym.getOutputSection() is nullptr. `ds->section->repl != ds->section`
-      // catches the ICF folded case. However, resolving a relocation in
-      // .debug_line to -1 would stop debugger users from setting breakpoints on
-      // the folded-in function, so exclude .debug_line.
+      // sym.getOutputSection() is nullptr. `ds->folded` catches the ICF folded
+      // case. However, resolving a relocation in .debug_line to -1 would stop
+      // debugger users from setting breakpoints on the folded-in function, so
+      // exclude .debug_line.
       //
       // For pre-DWARF-v5 .debug_loc and .debug_ranges, -1 is a reserved value
       // (base address selection entry), use 1 (which is used by GNU ld for
       // .debug_ranges).
       //
       // TODO To reduce disruption, we use 0 instead of -1 as the tombstone
       // value. Enable -1 in a future release.
       auto *ds = dyn_cast<Defined>(&sym);
-      if (!sym.getOutputSection() ||
-          (ds && ds->section->repl != ds->section && !isDebugLine)) {
+      if (!sym.getOutputSection() || (ds && ds->folded && !isDebugLine)) {
         // If -z dead-reloc-in-nonalloc= is specified, respect it.
         const uint64_t value = tombstone ? SignExtend64<bits>(*tombstone)
                                          : (isDebugLocOrRanges ? 1 : 0);
         target->relocateNoSym(bufLoc, type, value);
         continue;
       }
     }
     target->relocateNoSym(bufLoc, type, SignExtend64<bits>(sym.getVA(addend)));

lld/ELF/Relocations.cpp

std::pair<Thunk *, bool> ThunkCreator::getThunk(InputSection *isec,
 
   // We use a ((section, offset), addend) pair to find the thunk position if
   // possible so that we create only one thunk for aliased symbols or ICFed
   // sections. There may be multiple relocations sharing the same (section,
   // offset + addend) pair. We may revert the relocation back to its original
   // non-Thunk target, so we cannot fold offset + addend.
   if (auto *d = dyn_cast<Defined>(rel.sym))
     if (!d->isInPlt() && d->section)
-      thunkVec = &thunkedSymbolsBySectionAndAddend[{
-          {d->section->repl, d->value}, keyAddend}];
+      thunkVec = &thunkedSymbolsBySectionAndAddend[{{d->section, d->value},
+                                                    keyAddend}];
   if (!thunkVec)
     thunkVec = &thunkedSymbols[{rel.sym, keyAddend}];
 
   // Check existing Thunks for Sym to see if they can be reused
   for (Thunk *t : *thunkVec)
     if (isThunkSectionCompatible(isec, t->getThunkTargetSym()->section) &&
         t->isCompatibleWith(*isec, rel) &&
         target->inBranchRange(rel.type, src,

lld/ELF/Symbols.h

protected:
   Symbol(Kind k, InputFile *file, StringRefZ name, uint8_t binding,
          uint8_t stOther, uint8_t type)
       : file(file), nameData(name.data), nameSize(name.size), binding(binding),
         type(type), stOther(stOther), symbolKind(k), visibility(stOther & 3),
         isUsedInRegularObj(!file || file->kind() == InputFile::ObjKind),
         exportDynamic(isExportDynamic(k, visibility)), inDynamicList(false),
         canInline(false), referenced(false), traced(false), isInIplt(false),
         gotInIgot(false), isPreemptible(false), used(!config->gcSections),
-        needsTocRestore(false), scriptDefined(false), needsCopy(false),
-        needsGot(false), needsPlt(false), hasDirectReloc(false) {}
+        folded(false), needsTocRestore(false), scriptDefined(false),
+        needsCopy(false), needsGot(false), needsPlt(false),
+        hasDirectReloc(false) {}
 
 public:
   // True if this symbol is in the Iplt sub-section of the Plt and the Igot
   // sub-section of the .got.plt or .got.
   uint8_t isInIplt : 1;
 
   // True if this symbol needs a GOT entry and its GOT entry is actually in
   // Igot. This will be true only for certain non-preemptible ifuncs.
   uint8_t gotInIgot : 1;
 
   // True if this symbol is preemptible at load time.
   uint8_t isPreemptible : 1;
 
   // True if an undefined or shared symbol is used from a live section.
   //
   // NOTE: In Writer.cpp the field is used to mark local defined symbols
   // which are referenced by relocations when -r or --emit-relocs is given.
   uint8_t used : 1;
 
+  // True if defined relative to a section discarded by ICF.
+  uint8_t folded : 1;
+
   // True if a call to this symbol needs to be followed by a restore of the
   // PPC64 toc pointer.
   uint8_t needsTocRestore : 1;
 
   // True if this symbol is defined by a linker script.
   uint8_t scriptDefined : 1;
 
   // True if this symbol needs a canonical PLT entry, or (during

lld/ELF/Symbols.cpp

case Symbol::DefinedKind: {
     auto &d = cast<Defined>(sym);
     SectionBase *isec = d.section;
 
     // This is an absolute symbol.
     if (!isec)
       return d.value;
 
     assert(isec != &InputSection::discarded);
-    isec = isec->repl;
 
     uint64_t offset = d.value;
 
     // An object in an SHF_MERGE section might be referenced via a
     // section symbol (as a hack for reducing the number of local
     // symbols).
     // Depending on the addend, the reference via a section symbol
     // refers to a different object in the merge section.
void elf::maybeWarnUnorderableSymbol(const Symbol *sym) {
   if (sym->isUndefined())
     report(": unable to order undefined symbol: ");
   else if (sym->isShared())
     report(": unable to order shared symbol: ");
   else if (d && !d->section)
     report(": unable to order absolute symbol: ");
   else if (d && isa<OutputSection>(d->section))
     report(": unable to order synthetic symbol: ");
-  else if (d && !d->section->repl->isLive())
+  else if (d && !d->section->isLive())
     report(": unable to order discarded symbol: ");
 }
 
 // Returns true if a symbol can be replaced at load-time by a symbol
 // with the same name defined in other ELF executable or DSO.
 bool elf::computeIsPreemptible(const Symbol &sym) {
   assert(!sym.isLocal());
 

lld/ELF/SyntheticSections.cpp

if (firstRelI == (unsigned)-1)
     return nullptr;
 
   const RelTy &rel = rels[firstRelI];
   Symbol &b = sec->template getFile<ELFT>()->getRelocTargetSym(rel);
 
   // FDEs for garbage-collected or merged-by-ICF sections, or sections in
   // another partition, are dead.
   if (auto *d = dyn_cast<Defined>(&b))
-    if (d->section && d->section->partition == partition)
+    if (!d->folded && d->section && d->section->partition == partition)
       return d;
   return nullptr;
 }
 
 // .eh_frame is a sequence of CIE or FDE records. In general, there
 // is one CIE record per input object file which is followed by
 // a list of FDEs. This function searches an existing CIE or create a new
 // one and associates FDEs to the CIE.

lld/ELF/Writer.cpp

 }
 
 static bool includeInSymtab(const Symbol &b) {
   if (auto *d = dyn_cast<Defined>(&b)) {
     // Always include absolute symbols.
     SectionBase *sec = d->section;
     if (!sec)
       return true;
-    sec = sec->repl;
 
     // Exclude symbols pointing to garbage-collected sections.
     if (isa<InputSectionBase>(sec) && !sec->isLive())
       return false;
 
     if (auto *s = dyn_cast<MergeInputSection>(sec))
       if (!s->getSectionPiece(d->value)->live)
         return false;
if (it == symbolOrder.end())
       return;
     SymbolOrderEntry &ent = it->second;
     ent.present = true;
 
     maybeWarnUnorderableSymbol(&sym);
 
     if (auto *d = dyn_cast<Defined>(&sym)) {
       if (auto *sec = dyn_cast_or_null<InputSectionBase>(d->section)) {
-        int &priority = sectionOrder[cast<InputSectionBase>(sec->repl)];
+        int &priority = sectionOrder[cast<InputSectionBase>(sec)];
         priority = std::min(priority, ent.priority);
       }
     }
   };
 
   // We want both global and local symbols. We get the global ones from the
   // symbol table and iterate the object files for the local ones.
   for (Symbol *sym : symtab->symbols())
parallelForEach(File->getSymbols(), [&](Symbol *Sym) {
       auto *def = dyn_cast<Defined>(Sym);
       if (!def)
         return;
 
       const SectionBase *sec = def->section;
       if (!sec)
         return;
 
-      const InputSectionBase *inputSec = dyn_cast<InputSectionBase>(sec->repl);
+      const InputSectionBase *inputSec = dyn_cast<InputSectionBase>(sec);
       if (!inputSec || !inputSec->bytesDropped)
         return;
 
       const size_t OldSize = inputSec->data().size();
       const size_t NewSize = OldSize - inputSec->bytesDropped;
 
       if (def->value > NewSize && def->value <= OldSize) {
         LLVM_DEBUG(llvm::dbgs()