Merge branch 'main' into metadata/rootsignature1.2

Author: joaosaffran

.github/renovate.json

@@ -0,0 +1,12 @@
+{
+  "$schema": "https://docs.renovatebot.com/renovate-schema.json",
+  "extends": [
+    "config:recommended"
+  ],
+  "includePaths": [".github/**"],
+  "schedule": "* 0 * * 1",
+  "minimumReleaseAge": "3 days",
+  "assignees": ["boomanaiden154"],
+  "ignorePaths": [".github/workflows/containers/**"],
+  "groupName": "[Github] Update GHA Dependencies"
+}

.github/workflows/containers/github-action-ci/Dockerfile

@@ -2,7 +2,7 @@ FROM docker.io/library/ubuntu:24.04 as base
 ENV LLVM_SYSROOT=/opt/llvm
 
 FROM base as stage1-toolchain
-ENV LLVM_VERSION=21.1.0
+ENV LLVM_VERSION=21.1.1
 
 RUN apt-get update && \
     apt-get install -y \

.github/workflows/pr-code-lint.yml

@@ -33,7 +33,7 @@ jobs:
 
       - name: Get changed files
         id: changed-files
-        uses: step-security/changed-files@3dbe17c78367e7d60f00d78ae6781a35be47b4a1 # v45.0.1
+        uses: tj-actions/changed-files@ed68ef82c095e0d48ec87eccea555d944a631a4c # v46.0.5
         with:
           separator: ","
           skip_initial_fetch: true
@@ -47,10 +47,13 @@ jobs:
           echo "Changed files:"
           echo "$CHANGED_FILES"
 
+      # The clang tidy version should always be upgraded to the first version
+      # of a release cycle (x.1.0) or the last version of a release cycle, or
+      # if there have been relevant clang-format backports.
       - name: Install clang-tidy
         uses: aminya/setup-cpp@17c11551771948abc5752bbf3183482567c7caf0 # v1.1.1
         with:
-          clang-tidy: 20.1.8
+          clang-tidy: 21.1.0
 
       - name: Setup Python env
         uses: actions/setup-python@42375524e23c412d93fb67b49958b491fce71c38 # v5.4.0

bolt/include/bolt/Core/BinaryContext.h

@@ -326,6 +326,9 @@ class BinaryContext {
   /// Returns true if DWARF4 or lower is used.
   bool isDWARFLegacyUsed() const { return ContainsDwarfLegacy; }
 
+  /// Returns true if DWARFUnit is valid.
+  bool isValidDwarfUnit(DWARFUnit &DU) const;
+
   std::map<unsigned, DwarfLineTable> &getDwarfLineTables() {
     return DwarfLineTablesCUMap;
   }

bolt/include/bolt/Core/BinaryFunction.h

@@ -192,9 +192,6 @@ class BinaryFunction {
 
     mutable MCSymbol *FunctionConstantIslandLabel{nullptr};
     mutable MCSymbol *FunctionColdConstantIslandLabel{nullptr};
-
-    // Returns constant island alignment
-    uint16_t getAlignment() const { return sizeof(uint64_t); }
   };
 
   static constexpr uint64_t COUNT_NO_PROFILE =
@@ -2114,9 +2111,7 @@ class BinaryFunction {
     return *std::prev(CodeIter) <= *DataIter;
   }
 
-  uint16_t getConstantIslandAlignment() const {
-    return Islands ? Islands->getAlignment() : 1;
-  }
+  uint16_t getConstantIslandAlignment() const;
 
   /// If there is a constant island in the range [StartOffset, EndOffset),
   /// return its address.
@@ -2168,6 +2163,11 @@ class BinaryFunction {
     return Islands && !Islands->DataOffsets.empty();
   }
 
+  /// Return true if the whole function is a constant island.
+  bool isDataObject() const {
+    return Islands && Islands->CodeOffsets.size() == 0;
+  }
+
   bool isStartOfConstantIsland(uint64_t Offset) const {
     return hasConstantIsland() && Islands->DataOffsets.count(Offset);
   }

bolt/include/bolt/Core/DIEBuilder.h

@@ -217,7 +217,8 @@ class DIEBuilder {
                                std::optional<BOLTDWARF5AccelTableData *> Parent,
                                uint32_t NumberParentsInChain);
 
-  void registerUnit(DWARFUnit &DU, bool NeedSort);
+  /// Returns true if DWARFUnit is registered successfully.
+  bool registerUnit(DWARFUnit &DU, bool NeedSort);
 
   /// \return the unique ID of \p U if it exists.
   std::optional<uint32_t> getUnitId(const DWARFUnit &DU);

bolt/include/bolt/Core/FunctionLayout.h

@@ -232,20 +232,31 @@ class FunctionLayout {
     return Blocks[Index];
   }
 
+  /// Return the basic block after the given basic block iterator in the layout
+  /// or nullptr if the last basic block iterator is given.
+  const BinaryBasicBlock *getBasicBlockAfter(block_const_iterator BlockIt,
+                                             bool IgnoreSplits = true) const;
+
+  /// Returns the basic block after the given basic block in the layout or
+  /// nullptr if the last basic block is given.
+  ///
+  /// Note: prefer the version that takes the iterator as this function uses
+  /// linear basic block lookup.
+  const BinaryBasicBlock *getBasicBlockAfter(const BinaryBasicBlock *BB,
+                                             bool IgnoreSplits = true) const;
+
   /// Returns the basic block after the given basic block in the layout or
   /// nullptr if the last basic block is given.
+  ///
+  /// Note: prefer the version that takes the iterator as this function uses
+  /// linear basic block lookup.
   BinaryBasicBlock *getBasicBlockAfter(const BinaryBasicBlock *const BB,
                                        const bool IgnoreSplits = true) {
     return const_cast<BinaryBasicBlock *>(
         static_cast<const FunctionLayout &>(*this).getBasicBlockAfter(
             BB, IgnoreSplits));
   }
 
-  /// Returns the basic block after the given basic block in the layout or
-  /// nullptr if the last basic block is given.
-  const BinaryBasicBlock *getBasicBlockAfter(const BinaryBasicBlock *BB,
-                                             bool IgnoreSplits = true) const;
-
   /// True if the layout contains at least two non-empty fragments.
   bool isSplit() const;
 

bolt/include/bolt/Core/MCPlusBuilder.h

@@ -51,6 +51,7 @@ class raw_ostream;
 
 namespace bolt {
 class BinaryBasicBlock;
+class BinaryContext;
 class BinaryFunction;
 
 /// Different types of indirect branches encountered during disassembly.
@@ -530,10 +531,15 @@ class MCPlusBuilder {
     return 0;
   }
 
+  /// Create a helper function to increment counter for Instrumentation
+  virtual void createInstrCounterIncrFunc(BinaryContext &BC) {
+    llvm_unreachable("not implemented");
+  }
+
   /// Create increment contents of target by 1 for Instrumentation
-  virtual InstructionListType
-  createInstrIncMemory(const MCSymbol *Target, MCContext *Ctx, bool IsLeaf,
-                       unsigned CodePointerSize) const {
+  virtual InstructionListType createInstrIncMemory(const MCSymbol *Target,
+                                                   MCContext *Ctx, bool IsLeaf,
+                                                   unsigned CodePointerSize) {
     llvm_unreachable("not implemented");
     return InstructionListType();
   }

bolt/lib/Core/BinaryContext.cpp

@@ -1624,10 +1624,26 @@ DWARFContext *BinaryContext::getDWOContext() const {
   return &DWOCUs.begin()->second->getContext();
 }
 
+bool BinaryContext::isValidDwarfUnit(DWARFUnit &DU) const {
+  // Invalid DWARF unit with a DWOId but lacking a dwo_name.
+  if (DU.getDWOId() && !DU.isDWOUnit() &&
+      !DU.getUnitDIE().find(
+          {dwarf::DW_AT_dwo_name, dwarf::DW_AT_GNU_dwo_name})) {
+    this->outs() << "BOLT-ERROR: broken DWARF found in CU at offset 0x"
+                 << Twine::utohexstr(DU.getOffset()) << " (DWOId=0x"
+                 << Twine::utohexstr(*(DU.getDWOId()))
+                 << ", missing DW_AT_dwo_name / DW_AT_GNU_dwo_name)\n";
+    return false;
+  }
+  return true;
+}
+
 /// Handles DWO sections that can either be in .o, .dwo or .dwp files.
 void BinaryContext::preprocessDWODebugInfo() {
   for (const std::unique_ptr<DWARFUnit> &CU : DwCtx->compile_units()) {
     DWARFUnit *const DwarfUnit = CU.get();
+    if (!isValidDwarfUnit(*DwarfUnit))
+      continue;
     if (std::optional<uint64_t> DWOId = DwarfUnit->getDWOId()) {
       std::string DWOName = dwarf::toString(
           DwarfUnit->getUnitDIE().find(

bolt/lib/Core/BinaryFunction.cpp

@@ -284,6 +284,33 @@ BinaryFunction::getBasicBlockContainingOffset(uint64_t Offset) {
   return (Offset < BB->getOffset() + BB->getOriginalSize()) ? BB : nullptr;
 }
 
+uint16_t BinaryFunction::getConstantIslandAlignment() const {
+  if (Islands == nullptr)
+    return 1;
+
+  // For constant island inside a function, the default 8-byte alignment is
+  // probably good enough.
+  const uint16_t DefaultAlignment = sizeof(uint64_t);
+  if (!isDataObject())
+    return DefaultAlignment;
+
+  // If the constant island itself is a binary function, get its alignment
+  // based on its size, original address, and its owning section's alignment.
+  const uint64_t MaxAlignment =
+      std::min(uint64_t(1) << llvm::countr_zero(getAddress()),
+               OriginSection->getAlignment());
+  const uint64_t MinAlignment =
+      std::max((uint64_t)DefaultAlignment,
+               uint64_t(1) << (63 - llvm::countl_zero(getSize())));
+  uint64_t Alignment = std::min(MinAlignment, MaxAlignment);
+  if (Alignment >> 16) {
+    BC.errs() << "BOLT-ERROR: the constant island's alignment is too big: 0x"
+              << Twine::utohexstr(Alignment) << "\n";
+    exit(1);
+  }
+  return (uint16_t)Alignment;
+}
+
 void BinaryFunction::markUnreachableBlocks() {
   std::stack<BinaryBasicBlock *> Stack;
 
@@ -3598,7 +3625,9 @@ void BinaryFunction::fixBranches() {
   auto &MIB = BC.MIB;
   MCContext *Ctx = BC.Ctx.get();
 
-  for (BinaryBasicBlock *BB : BasicBlocks) {
+  for (auto BBI = Layout.block_begin(), BBE = Layout.block_end(); BBI != BBE;
+       ++BBI) {
+    BinaryBasicBlock *BB = *BBI;
     const MCSymbol *TBB = nullptr;
     const MCSymbol *FBB = nullptr;
     MCInst *CondBranch = nullptr;
@@ -3612,7 +3641,7 @@ void BinaryFunction::fixBranches() {
 
     // Basic block that follows the current one in the final layout.
     const BinaryBasicBlock *const NextBB =
-        Layout.getBasicBlockAfter(BB, /*IgnoreSplits=*/false);
+        Layout.getBasicBlockAfter(BBI, /*IgnoreSplits*/ false);
 
     if (BB->succ_size() == 1) {
       // __builtin_unreachable() could create a conditional branch that