Merge branch 'llvm:main' into counted-by-on-struct-pointers

Author: bwendling

.clang-format-ignore

@@ -19,13 +19,13 @@ jobs:
         include:
           - os: ubuntu-24.04
             ccache-variant: sccache
-            c_compiler: clang
-            cpp_compiler: clang++
+            c_compiler: clang-20
+            cpp_compiler: clang++-20
           # TODO: remove ccache logic when https://github.com/hendrikmuhs/ccache-action/issues/279 is resolved.
           - os: ubuntu-24.04-arm
             ccache-variant: ccache
-            c_compiler: clang
-            cpp_compiler: clang++
+            c_compiler: clang-20
+            cpp_compiler: clang++-20
           # TODO: add back gcc build when it is fixed
           # - c_compiler: gcc
           #   cpp_compiler: g++
@@ -51,6 +51,9 @@ jobs:
     #   For more information, see https://wiki.debian.org/Multiarch/LibraryPathOverview
     - name: Prepare dependencies (Ubuntu)
       run: |
+        wget https://apt.llvm.org/llvm.sh
+        chmod +x llvm.sh
+        sudo ./llvm.sh 20
         sudo apt-get update
         sudo apt-get install -y libmpfr-dev libgmp-dev libmpc-dev ninja-build linux-libc-dev
         sudo ln -sf /usr/include/$(uname -p)-linux-gnu/asm /usr/include/asm

.github/workflows/libc-fullbuild-tests.yml

@@ -69,7 +69,7 @@ jobs:
       - name: Upload Artifacts
         uses: actions/upload-artifact@65c4c4a1ddee5b72f698fdd19549f0f0fb45cf08 # v4.6.0
         with:
-          name: Premerge Artifacts
+          name: Premerge Artifacts (Linux)
           path: artifacts/
           retention-days: 5
           include-hidden-files: 'true'
@@ -124,7 +124,7 @@ jobs:
       - name: Upload Artifacts
         uses: actions/upload-artifact@65c4c4a1ddee5b72f698fdd19549f0f0fb45cf08 # v4.6.0
         with:
-          name: Premerge Artifacts
+          name: Premerge Artifacts (Windows)
           path: artifacts/
           retention-days: 5
           include-hidden-files: 'true'

.github/workflows/premerge.yaml

@@ -82,7 +82,8 @@ endforeach()
 
 set(BOLT_ENABLE_RUNTIME_default OFF)
 if ((CMAKE_SYSTEM_PROCESSOR STREQUAL "x86_64"
-    OR CMAKE_SYSTEM_PROCESSOR MATCHES "^(arm64|aarch64)$")
+    OR CMAKE_SYSTEM_PROCESSOR MATCHES "^(arm64|aarch64)$"
+    OR CMAKE_SYSTEM_PROCESSOR STREQUAL "riscv64")
     AND (CMAKE_SYSTEM_NAME STREQUAL "Linux"
       OR CMAKE_SYSTEM_NAME STREQUAL "Darwin")
     AND (NOT CMAKE_CROSSCOMPILING))
@@ -136,7 +137,7 @@ if (LLVM_INCLUDE_TESTS)
 endif()
 
 if (BOLT_ENABLE_RUNTIME)
-  message(STATUS "Building BOLT runtime libraries for X86")
+  message(STATUS "Building BOLT runtime libraries for ${CMAKE_SYSTEM_PROCESSOR}")
   set(extra_args "")
   if(CMAKE_SYSROOT)
     list(APPEND extra_args -DCMAKE_SYSROOT=${CMAKE_SYSROOT})

bolt/CMakeLists.txt

@@ -1174,6 +1174,11 @@ class BinaryFunction {
     return getSecondaryEntryPointSymbol(BB.getLabel());
   }
 
+  /// Remove a label from the secondary entry point map.
+  void removeSymbolFromSecondaryEntryPointMap(const MCSymbol *Label) {
+    SecondaryEntryPoints.erase(Label);
+  }
+
   /// Return true if the basic block is an entry point into the function
   /// (either primary or secondary).
   bool isEntryPoint(const BinaryBasicBlock &BB) const {
@@ -2126,6 +2131,10 @@ class BinaryFunction {
     return Islands && !Islands->DataOffsets.empty();
   }
 
+  bool isStartOfConstantIsland(uint64_t Offset) const {
+    return hasConstantIsland() && Islands->DataOffsets.count(Offset);
+  }
+
   /// Return true iff the symbol could be seen inside this function otherwise
   /// it is probably another function.
   bool isSymbolValidInScope(const SymbolRef &Symbol, uint64_t SymbolSize) const;

bolt/include/bolt/Core/BinaryFunction.h

@@ -1896,6 +1896,15 @@ void BinaryFunction::postProcessEntryPoints() {
     if (BC.isAArch64() && Offset == getSize())
       continue;
 
+    // If we have grabbed a wrong code label which actually points to some
+    // constant island inside the function, ignore this label and remove it
+    // from the secondary entry point map.
+    if (isStartOfConstantIsland(Offset)) {
+      BC.SymbolToFunctionMap.erase(Label);
+      removeSymbolFromSecondaryEntryPointMap(Label);
+      continue;
+    }
+
     BC.errs() << "BOLT-WARNING: reference in the middle of instruction "
                  "detected in function "
               << *this << " at offset 0x" << Twine::utohexstr(Offset) << '\n';

bolt/lib/Core/BinaryFunction.cpp

@@ -123,6 +123,7 @@ static bool isSupportedRISCV(uint32_t Type) {
   case ELF::R_RISCV_LO12_S:
   case ELF::R_RISCV_64:
   case ELF::R_RISCV_TLS_GOT_HI20:
+  case ELF::R_RISCV_TLS_GD_HI20:
   case ELF::R_RISCV_TPREL_HI20:
   case ELF::R_RISCV_TPREL_ADD:
   case ELF::R_RISCV_TPREL_LO12_I:
@@ -236,6 +237,7 @@ static size_t getSizeForTypeRISCV(uint32_t Type) {
   case ELF::R_RISCV_64:
   case ELF::R_RISCV_GOT_HI20:
   case ELF::R_RISCV_TLS_GOT_HI20:
+  case ELF::R_RISCV_TLS_GD_HI20:
     // See extractValueRISCV for why this is necessary.
     return 8;
   }
@@ -491,6 +493,7 @@ static uint64_t extractValueRISCV(uint32_t Type, uint64_t Contents,
     return extractBImmRISCV(Contents);
   case ELF::R_RISCV_GOT_HI20:
   case ELF::R_RISCV_TLS_GOT_HI20:
+  case ELF::R_RISCV_TLS_GD_HI20:
     // We need to know the exact address of the GOT entry so we extract the
     // value from both the AUIPC and L[D|W]. We cannot rely on the symbol in the
     // relocation for this since it simply refers to the object that is stored
@@ -707,6 +710,7 @@ static bool isPCRelativeRISCV(uint32_t Type) {
   case ELF::R_RISCV_RVC_BRANCH:
   case ELF::R_RISCV_32_PCREL:
   case ELF::R_RISCV_TLS_GOT_HI20:
+  case ELF::R_RISCV_TLS_GD_HI20:
     return true;
   }
 }

bolt/lib/Core/Relocation.cpp

@@ -597,8 +597,9 @@ Error RewriteInstance::discoverStorage() {
 
   // Hugify: Additional huge page from left side due to
   // weird ASLR mapping addresses (4KB aligned)
-  if (opts::Hugify && !BC->HasFixedLoadAddress)
+  if (opts::Hugify && !BC->HasFixedLoadAddress) {
     NextAvailableAddress += BC->PageAlign;
+  }
 
   if (!opts::UseGnuStack && !BC->IsLinuxKernel) {
     // This is where the black magic happens. Creating PHDR table in a segment
@@ -2925,12 +2926,12 @@ void RewriteInstance::handleRelocation(const SectionRef &RelocatedSection,
 
     if (BinaryData *BD = BC->getBinaryDataContainingAddress(SymbolAddress)) {
       // Note: this assertion is trying to check sanity of BinaryData objects
-      // but AArch64 has inferred and incomplete object locations coming from
-      // GOT/TLS or any other non-trivial relocation (that requires creation
-      // of sections and whose symbol address is not really what should be
-      // encoded in the instruction). So we essentially disabled this check
+      // but AArch64 and RISCV has inferred and incomplete object locations
+      // coming from GOT/TLS or any other non-trivial relocation (that requires
+      // creation of sections and whose symbol address is not really what should
+      // be encoded in the instruction). So we essentially disabled this check
       // for AArch64 and live with bogus names for objects.
-      assert((IsAArch64 || IsSectionRelocation ||
+      assert((IsAArch64 || BC->isRISCV() || IsSectionRelocation ||
               BD->nameStartsWith(SymbolName) ||
               BD->nameStartsWith("PG" + SymbolName) ||
               (BD->nameStartsWith("ANONYMOUS") &&
@@ -5885,17 +5886,28 @@ void RewriteInstance::rewriteFile() {
 
   // Write all allocatable sections - reloc-mode text is written here as well
   for (BinarySection &Section : BC->allocatableSections()) {
-    if (!Section.isFinalized() || !Section.getOutputData())
+    if (!Section.isFinalized() || !Section.getOutputData()) {
+      LLVM_DEBUG(if (opts::Verbosity > 1) {
+        dbgs() << "BOLT-INFO: new section is finalized or !getOutputData, skip "
+               << Section.getName() << '\n';
+      });
       continue;
-    if (Section.isLinkOnly())
+    }
+    if (Section.isLinkOnly()) {
+      LLVM_DEBUG(if (opts::Verbosity > 1) {
+        dbgs() << "BOLT-INFO: new section is link only, skip "
+               << Section.getName() << '\n';
+      });
       continue;
+    }
 
     if (opts::Verbosity >= 1)
       BC->outs() << "BOLT: writing new section " << Section.getName()
                  << "\n data at 0x"
                  << Twine::utohexstr(Section.getAllocAddress()) << "\n of size "
                  << Section.getOutputSize() << "\n at offset "
-                 << Section.getOutputFileOffset() << '\n';
+                 << Section.getOutputFileOffset() << " with content size "
+                 << Section.getOutputContents().size() << '\n';
     OS.seek(Section.getOutputFileOffset());
     Section.write(OS);
   }