Merge branch 'llvm:main' into llvmgh-101657

.github/CODEOWNERS

@@ -33,9 +33,6 @@
 /clang/www/cxx_dr_status.html @Endilll
 /clang/www/make_cxx_dr_status @Endilll
 
-clang/lib/AST/Interp/ @tbaederr
-clang/test/AST/Interp/ @tbaederr
-
 /clang/include/clang/CIR @lanza @bcardosolopes
 /clang/lib/CIR @lanza @bcardosolopes
 /clang/tools/cir-* @lanza @bcardosolopes

.github/workflows/libclang-python-tests.yml

@@ -22,12 +22,6 @@ on:
       - '.github/workflows/libclang-python-tests.yml'
       - '.github/workflows/llvm-project-tests.yml'
 
-concurrency:
-  # Skip intermediate builds: always.
-  # Cancel intermediate builds: only if it is a pull request build.
-  group: ${{ github.workflow }}-${{ github.ref }}
-  cancel-in-progress: ${{ startsWith(github.ref, 'refs/pull/') }}
-
 jobs:
   check-clang-python:
     # Build libclang and then run the libclang Python binding's unit tests.

.github/workflows/llvm-project-tests.yml

@@ -51,7 +51,7 @@ concurrency:
   # Cancel intermediate builds: only if it is a pull request build.
   # If the group name here is the same as the group name in the workflow that includes
   # this one, then the action will try to wait on itself and get stuck.
-  group: llvm-project-${{ github.workflow }}-${{ inputs.projects }}${{ github.ref }}
+  group: llvm-project-${{ github.workflow }}-${{ inputs.projects }}-${{ inputs.python_version }}${{ github.ref }}
   cancel-in-progress: ${{ startsWith(github.ref, 'refs/pull/') }}
 
 jobs:

.github/workflows/release-tasks.yml

@@ -101,6 +101,7 @@ jobs:
   release-sources:
     name: Package Release Sources
     permissions:
+      contents: read
       id-token: write
       attestations: write
     needs:

bolt/CMakeLists.txt

@@ -82,7 +82,7 @@ endforeach()
 
 set(BOLT_ENABLE_RUNTIME_default OFF)
 if ((CMAKE_SYSTEM_PROCESSOR STREQUAL "x86_64"
-    OR CMAKE_SYSTEM_PROCESSOR STREQUAL "aarch64")
+    OR CMAKE_SYSTEM_PROCESSOR MATCHES "^(arm64|aarch64)$")
     AND (CMAKE_SYSTEM_NAME STREQUAL "Linux"
       OR CMAKE_SYSTEM_NAME STREQUAL "Darwin")
     AND (NOT CMAKE_CROSSCOMPILING))

bolt/docs/CommandLineArgumentReference.md

@@ -113,6 +113,10 @@
 
   Prints out offsets for abbrev and debug_info of Skeleton CUs that get patched.
 
+- `--debug-thread-count=<uint>`
+
+  Specifies the number of threads to be used when processing DWO debug information.
+
 - `--dot-tooltip-code`
 
   Add basic block instructions as tool tips on nodes

bolt/docs/OptimizingLinux.md

@@ -37,7 +37,7 @@ Convert `perf` profile into a format suitable for BOLT passing the `vmlinux` bin
 
 
 ```bash
-$ sudo chwon $USER perf.data
+$ sudo chown $USER perf.data
 $ perf2bolt -p perf.data -o perf.fdata vmlinux
 ```
 

bolt/include/bolt/Core/ParallelUtilities.h

@@ -50,7 +50,8 @@ enum SchedulingPolicy {
 };
 
 /// Return the managed thread pool and initialize it if not initialized.
-ThreadPoolInterface &getThreadPool();
+ThreadPoolInterface &
+getThreadPool(const unsigned ThreadsCount = opts::ThreadCount);
 
 /// Perform the work on each BinaryFunction except those that are accepted
 /// by SkipPredicate, scheduling heuristic is based on SchedPolicy.

bolt/include/bolt/Rewrite/DWARFRewriter.h

@@ -184,7 +184,8 @@ class DWARFRewriter {
   /// Output .dwo files.
   void writeDWOFiles(DWARFUnit &, const OverriddenSectionsMap &,
                      const std::string &, DebugLocWriter &,
-                     DebugStrOffsetsWriter &, DebugStrWriter &);
+                     DebugStrOffsetsWriter &, DebugStrWriter &,
+                     DebugRangesSectionWriter &);
   using KnownSectionsEntry = std::pair<MCSection *, DWARFSectionKind>;
 };
 

bolt/lib/Core/BinaryContext.cpp

@@ -142,7 +142,6 @@ BinaryContext::BinaryContext(std::unique_ptr<MCContext> Ctx,
       InstPrinter(std::move(InstPrinter)), MIA(std::move(MIA)),
       MIB(std::move(MIB)), MRI(std::move(MRI)), DisAsm(std::move(DisAsm)),
       Logger(Logger), InitialDynoStats(isAArch64()) {
-  Relocation::Arch = this->TheTriple->getArch();
   RegularPageSize = isAArch64() ? RegularPageSizeAArch64 : RegularPageSizeX86;
   PageAlign = opts::NoHugePages ? RegularPageSize : HugePageSize;
 }

bolt/lib/Core/ParallelUtilities.cpp

@@ -49,7 +49,7 @@ namespace ParallelUtilities {
 
 namespace {
 /// A single thread pool that is used to run parallel tasks
-std::unique_ptr<DefaultThreadPool> ThreadPoolPtr;
+std::unique_ptr<ThreadPoolInterface> ThreadPoolPtr;
 
 unsigned computeCostFor(const BinaryFunction &BF,
                         const PredicateTy &SkipPredicate,
@@ -102,12 +102,15 @@ inline unsigned estimateTotalCost(const BinaryContext &BC,
 
 } // namespace
 
-ThreadPoolInterface &getThreadPool() {
+ThreadPoolInterface &getThreadPool(const unsigned ThreadsCount) {
   if (ThreadPoolPtr.get())
     return *ThreadPoolPtr;
 
-  ThreadPoolPtr = std::make_unique<DefaultThreadPool>(
-      llvm::hardware_concurrency(opts::ThreadCount));
+  if (ThreadsCount > 1)
+    ThreadPoolPtr = std::make_unique<DefaultThreadPool>(
+        llvm::hardware_concurrency(ThreadsCount));
+  else
+    ThreadPoolPtr = std::make_unique<SingleThreadExecutor>();
   return *ThreadPoolPtr;
 }
 

bolt/lib/Rewrite/DWARFRewriter.cpp

@@ -329,6 +329,12 @@ static cl::opt<bool> KeepARanges(
         "keep or generate .debug_aranges section if .gdb_index is written"),
     cl::Hidden, cl::cat(BoltCategory));
 
+static cl::opt<unsigned>
+    DebugThreadCount("debug-thread-count",
+                     cl::desc("specifies thread count for the multithreading "
+                              "for updating DWO debug info"),
+                     cl::init(1), cl::cat(BoltCategory));
+
 static cl::opt<std::string> DwarfOutputPath(
     "dwarf-output-path",
     cl::desc("Path to where .dwo files will be written out to."), cl::init(""),
@@ -475,8 +481,8 @@ static void emitDWOBuilder(const std::string &DWOName,
                            DWARFUnit &SplitCU, DWARFUnit &CU,
                            DebugLocWriter &LocWriter,
                            DebugStrOffsetsWriter &StrOffstsWriter,
-                           DebugStrWriter &StrWriter,
-                           GDBIndex &GDBIndexSection) {
+                           DebugStrWriter &StrWriter, GDBIndex &GDBIndexSection,
+                           DebugRangesSectionWriter &TempRangesSectionWriter) {
   // Populate debug_info and debug_abbrev for current dwo into StringRef.
   DWODIEBuilder.generateAbbrevs();
   DWODIEBuilder.finish();
@@ -532,7 +538,7 @@ static void emitDWOBuilder(const std::string &DWOName,
     OverriddenSections[Kind] = Contents;
   }
   Rewriter.writeDWOFiles(CU, OverriddenSections, DWOName, LocWriter,
-                         StrOffstsWriter, StrWriter);
+                         StrOffstsWriter, StrWriter, TempRangesSectionWriter);
 }
 
 using DWARFUnitVec = std::vector<DWARFUnit *>;
@@ -646,7 +652,6 @@ void DWARFRewriter::updateDebugInfo() {
                                          *StrWriter);
   GDBIndex GDBIndexSection(BC);
   auto processSplitCU = [&](DWARFUnit &Unit, DWARFUnit &SplitCU,
-                            DIEBuilder &DIEBlder,
                             DebugRangesSectionWriter &TempRangesSectionWriter,
                             DebugAddrWriter &AddressWriter,
                             const std::string &DWOName,
@@ -669,7 +674,7 @@ void DWARFRewriter::updateDebugInfo() {
 
     emitDWOBuilder(DWOName, DWODIEBuilder, *this, SplitCU, Unit,
                    DebugLocDWoWriter, DWOStrOffstsWriter, DWOStrWriter,
-                   GDBIndexSection);
+                   GDBIndexSection, TempRangesSectionWriter);
   };
   auto processMainBinaryCU = [&](DWARFUnit &Unit, DIEBuilder &DIEBlder) {
     std::optional<DWARFUnit *> SplitCU;
@@ -716,9 +721,13 @@ void DWARFRewriter::updateDebugInfo() {
       finalizeTypeSections(DIEBlder, *Streamer, GDBIndexSection);
 
   CUPartitionVector PartVec = partitionCUs(*BC.DwCtx);
+  const unsigned int ThreadCount =
+      std::min(opts::DebugThreadCount, opts::ThreadCount);
   for (std::vector<DWARFUnit *> &Vec : PartVec) {
     DIEBlder.buildCompileUnits(Vec);
     llvm::SmallVector<std::unique_ptr<DIEBuilder>, 72> DWODIEBuildersByCU;
+    ThreadPoolInterface &ThreadPool =
+        ParallelUtilities::getThreadPool(ThreadCount);
     for (DWARFUnit *CU : DIEBlder.getProcessedCUs()) {
       createRangeLocListAddressWriters(*CU);
       std::optional<DWARFUnit *> SplitCU;
@@ -729,9 +738,9 @@ void DWARFRewriter::updateDebugInfo() {
         continue;
       DebugAddrWriter &AddressWriter =
           *AddressWritersByCU[CU->getOffset()].get();
-      DebugRangesSectionWriter *TempRangesSectionWriter =
-          CU->getVersion() >= 5 ? RangeListsWritersByCU[*DWOId].get()
-                                : LegacyRangesWritersByCU[*DWOId].get();
+      DebugRangesSectionWriter &TempRangesSectionWriter =
+          CU->getVersion() >= 5 ? *RangeListsWritersByCU[*DWOId].get()
+                                : *LegacyRangesWritersByCU[*DWOId].get();
       std::optional<std::string> DwarfOutputPath =
           opts::DwarfOutputPath.empty()
               ? std::nullopt
@@ -744,9 +753,17 @@ void DWARFRewriter::updateDebugInfo() {
           *DWODIEBuildersByCU.emplace_back(std::move(DWODIEBuilderPtr)).get();
       if (CU->getVersion() >= 5)
         StrOffstsWriter->finalizeSection(*CU, DIEBlder);
-      processSplitCU(*CU, **SplitCU, DIEBlder, *TempRangesSectionWriter,
-                     AddressWriter, DWOName, DwarfOutputPath, DWODIEBuilder);
+      // Important to capture CU and SplitCU by value here, otherwise when the
+      // thread is executed at some point after the current iteration of the
+      // loop, dereferencing CU/SplitCU in the call to processSplitCU means it
+      // will dereference a different variable than the one intended, causing a
+      // seg fault.
+      ThreadPool.async([&, DwarfOutputPath, DWOName, CU, SplitCU] {
+        processSplitCU(*CU, **SplitCU, TempRangesSectionWriter, AddressWriter,
+                       DWOName, DwarfOutputPath, DWODIEBuilder);
+      });
     }
+    ThreadPool.wait();
     for (std::unique_ptr<DIEBuilder> &DWODIEBuilderPtr : DWODIEBuildersByCU)
       DWODIEBuilderPtr->updateDebugNamesTable();
     for (DWARFUnit *CU : DIEBlder.getProcessedCUs())
@@ -1807,7 +1824,8 @@ std::optional<StringRef> updateDebugData(
 void DWARFRewriter::writeDWOFiles(
     DWARFUnit &CU, const OverriddenSectionsMap &OverridenSections,
     const std::string &DWOName, DebugLocWriter &LocWriter,
-    DebugStrOffsetsWriter &StrOffstsWriter, DebugStrWriter &StrWriter) {
+    DebugStrOffsetsWriter &StrOffstsWriter, DebugStrWriter &StrWriter,
+    DebugRangesSectionWriter &TempRangesSectionWriter) {
   // Setup DWP code once.
   DWARFContext *DWOCtx = BC.getDWOContext();
   const uint64_t DWOId = *CU.getDWOId();
@@ -1854,9 +1872,8 @@ void DWARFRewriter::writeDWOFiles(
 
   DebugRangeListsSectionWriter *RangeListssWriter = nullptr;
   if (CU.getVersion() == 5) {
-    assert(RangeListsWritersByCU.count(DWOId) != 0 &&
-           "No RangeListsWriter for DWO ID.");
-    RangeListssWriter = RangeListsWritersByCU[DWOId].get();
+    RangeListssWriter =
+        llvm::dyn_cast<DebugRangeListsSectionWriter>(&TempRangesSectionWriter);
 
     // Handling .debug_rnglists.dwo separately. The original .o/.dwo might not
     // have .debug_rnglists so won't be part of the loop below.

bolt/lib/Rewrite/MachORewriteInstance.cpp

@@ -72,6 +72,7 @@ MachORewriteInstance::MachORewriteInstance(object::MachOObjectFile *InputFile,
                                            StringRef ToolPath, Error &Err)
     : InputFile(InputFile), ToolPath(ToolPath) {
   ErrorAsOutParameter EAO(&Err);
+  Relocation::Arch = InputFile->makeTriple().getArch();
   auto BCOrErr = BinaryContext::createBinaryContext(
       InputFile->makeTriple(), InputFile->getFileName(), nullptr,
       /* IsPIC */ true, DWARFContext::create(*InputFile),

bolt/lib/Rewrite/RewriteInstance.cpp

@@ -354,6 +354,7 @@ RewriteInstance::RewriteInstance(ELFObjectFileBase *File, const int Argc,
     }
   }
 
+  Relocation::Arch = TheTriple.getArch();
   auto BCOrErr = BinaryContext::createBinaryContext(
       TheTriple, File->getFileName(), Features.get(), IsPIC,
       DWARFContext::create(*File, DWARFContext::ProcessDebugRelocations::Ignore,
@@ -2612,7 +2613,7 @@ void RewriteInstance::handleRelocation(const SectionRef &RelocatedSection,
       Expected<StringRef> SectionName = Section->getName();
       if (SectionName && !SectionName->empty())
         ReferencedSection = BC->getUniqueSectionByName(*SectionName);
-    } else if (ReferencedSymbol && ContainingBF &&
+    } else if (BC->isRISCV() && ReferencedSymbol && ContainingBF &&
                (cantFail(Symbol.getFlags()) & SymbolRef::SF_Absolute)) {
       // This might be a relocation for an ABS symbols like __global_pointer$ on
       // RISC-V

bolt/test/AArch64/Inputs/build_id.ldscript

@@ -0,0 +1,9 @@
+SECTIONS
+{
+  PROVIDE (__executable_start = SEGMENT_START("text-segment", 0x400000)); . = SEGMENT_START("text-segment", 0x400000) + SIZEOF_HEADERS;
+  .note.gnu.build-id (0x400400):
+   {
+    build_id_note = ABSOLUTE(.);
+    *(.note.gnu.build-id)
+   }
+}

bolt/test/AArch64/build_id.c

@@ -0,0 +1,25 @@
+// This test checks that referencing build_id through GOT table
+// would result in GOT access after disassembly, not directly
+// to build_id address.
+
+// RUN: %clang %cflags -fuse-ld=lld -Wl,-T,%S/Inputs/build_id.ldscript -Wl,-q \
+// RUN:   -Wl,--no-relax -Wl,--build-id=sha1 %s -o %t.exe
+// RUN: llvm-bolt -print-disasm --print-only=get_build_id %t.exe -o %t.bolt | \
+// RUN:   FileCheck %s
+
+// CHECK: adrp	[[REG:x[0-28]+]], __BOLT_got_zero
+// CHECK: ldr x{{.*}}, [[[REG]], :lo12:__BOLT_got_zero{{.*}}]
+
+struct build_id_note {
+  char pad[16];
+  char hash[20];
+};
+
+extern const struct build_id_note build_id_note;
+
+__attribute__((noinline)) char get_build_id() { return build_id_note.hash[0]; }
+
+int main() {
+  get_build_id();
+  return 0;
+}