Merge branch 'main' into fixdistinct

Author: rbintel

bolt/include/bolt/Profile/DataAggregator.h

@@ -170,6 +170,9 @@ class DataAggregator : public DataReader {
   std::string BuildIDBinaryName;
 
   /// Memory map info for a single file as recorded in perf.data
+  /// When a binary has multiple text segments, the Size is computed as the
+  /// difference of the last address of these segments from the BaseAddress.
+  /// The base addresses of all text segments must be the same.
   struct MMapInfo {
     uint64_t BaseAddress{0}; /// Base address of the mapped binary.
     uint64_t MMapAddress{0}; /// Address of the executable segment.
@@ -493,6 +496,11 @@ class DataAggregator : public DataReader {
   /// and return a file name matching a given \p FileBuildID.
   std::optional<StringRef> getFileNameForBuildID(StringRef FileBuildID);
 
+  /// Get a constant reference to the parsed binary mmap entries.
+  const std::unordered_map<uint64_t, MMapInfo> &getBinaryMMapInfo() {
+    return BinaryMMapInfo;
+  }
+
   friend class YAMLProfileWriter;
 };
 } // namespace bolt

bolt/lib/Profile/DataAggregator.cpp

@@ -95,6 +95,12 @@ cl::opt<bool> ReadPreAggregated(
     "pa", cl::desc("skip perf and read data from a pre-aggregated file format"),
     cl::cat(AggregatorCategory));
 
+cl::opt<std::string>
+    ReadPerfEvents("perf-script-events",
+                   cl::desc("skip perf event collection by supplying a "
+                            "perf-script output in a textual format"),
+                   cl::ReallyHidden, cl::init(""), cl::cat(AggregatorCategory));
+
 static cl::opt<bool>
 TimeAggregator("time-aggr",
   cl::desc("time BOLT aggregator"),
@@ -167,8 +173,9 @@ void DataAggregator::findPerfExecutable() {
 void DataAggregator::start() {
   outs() << "PERF2BOLT: Starting data aggregation job for " << Filename << "\n";
 
-  // Don't launch perf for pre-aggregated files
-  if (opts::ReadPreAggregated)
+  // Don't launch perf for pre-aggregated files or when perf input is specified
+  // by the user.
+  if (opts::ReadPreAggregated || !opts::ReadPerfEvents.empty())
     return;
 
   findPerfExecutable();
@@ -464,6 +471,13 @@ void DataAggregator::filterBinaryMMapInfo() {
 
 int DataAggregator::prepareToParse(StringRef Name, PerfProcessInfo &Process,
                                    PerfProcessErrorCallbackTy Callback) {
+  if (!opts::ReadPerfEvents.empty()) {
+    outs() << "PERF2BOLT: using pre-processed perf events for '" << Name
+           << "' (perf-script-events)\n";
+    ParsingBuf = opts::ReadPerfEvents;
+    return 0;
+  }
+
   std::string Error;
   outs() << "PERF2BOLT: waiting for perf " << Name
          << " collection to finish...\n";
@@ -2056,15 +2070,6 @@ std::error_code DataAggregator::parseMMapEvents() {
     if (FileMMapInfo.first == "(deleted)")
       continue;
 
-    // Consider only the first mapping of the file for any given PID
-    auto Range = GlobalMMapInfo.equal_range(FileMMapInfo.first);
-    bool PIDExists = llvm::any_of(make_range(Range), [&](const auto &MI) {
-      return MI.second.PID == FileMMapInfo.second.PID;
-    });
-
-    if (PIDExists)
-      continue;
-
     GlobalMMapInfo.insert(FileMMapInfo);
   }
 
@@ -2116,12 +2121,22 @@ std::error_code DataAggregator::parseMMapEvents() {
                << " using file offset 0x" << Twine::utohexstr(MMapInfo.Offset)
                << ". Ignoring profile data for this mapping\n";
         continue;
-      } else {
-        MMapInfo.BaseAddress = *BaseAddress;
       }
+      MMapInfo.BaseAddress = *BaseAddress;
     }
 
-    BinaryMMapInfo.insert(std::make_pair(MMapInfo.PID, MMapInfo));
+    // Try to add MMapInfo to the map and update its size. Large binaries may
+    // span to multiple text segments, so the mapping is inserted only on the
+    // first occurrence.
+    if (!BinaryMMapInfo.insert(std::make_pair(MMapInfo.PID, MMapInfo)).second)
+      assert(MMapInfo.BaseAddress == BinaryMMapInfo[MMapInfo.PID].BaseAddress &&
+             "Base address on multiple segment mappings should match");
+
+    // Update mapping size.
+    const uint64_t EndAddress = MMapInfo.MMapAddress + MMapInfo.Size;
+    const uint64_t Size = EndAddress - BinaryMMapInfo[MMapInfo.PID].BaseAddress;
+    if (Size > BinaryMMapInfo[MMapInfo.PID].Size)
+      BinaryMMapInfo[MMapInfo.PID].Size = Size;
   }
 
   if (BinaryMMapInfo.empty()) {

bolt/unittests/Core/CMakeLists.txt

@@ -8,6 +8,7 @@ set(LLVM_LINK_COMPONENTS
 add_bolt_unittest(CoreTests
   BinaryContext.cpp
   MCPlusBuilder.cpp
+  MemoryMaps.cpp
   DynoStats.cpp
 
   DISABLE_LLVM_LINK_LLVM_DYLIB
@@ -17,6 +18,8 @@ target_link_libraries(CoreTests
   PRIVATE
   LLVMBOLTCore
   LLVMBOLTRewrite
+  LLVMBOLTProfile
+  LLVMTestingSupport
   )
 
 foreach (tgt ${BOLT_TARGETS_TO_BUILD})

bolt/unittests/Core/MemoryMaps.cpp

@@ -0,0 +1,142 @@
+//===- bolt/unittest/Core/MemoryMaps.cpp ----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include "bolt/Core/BinaryContext.h"
+#include "bolt/Profile/DataAggregator.h"
+#include "llvm/BinaryFormat/ELF.h"
+#include "llvm/DebugInfo/DWARF/DWARFContext.h"
+#include "llvm/Support/CommandLine.h"
+#include "llvm/Support/TargetSelect.h"
+#include "llvm/Testing/Support/Error.h"
+#include "gtest/gtest.h"
+
+using namespace llvm;
+using namespace llvm::object;
+using namespace llvm::ELF;
+using namespace bolt;
+
+namespace opts {
+extern cl::opt<std::string> ReadPerfEvents;
+} // namespace opts
+
+namespace {
+
+/// Perform checks on memory map events normally captured in perf. Tests use
+/// the 'opts::ReadPerfEvents' flag to emulate these events, passing a custom
+/// 'perf script' output to DataAggregator.
+struct MemoryMapsTester : public testing::TestWithParam<Triple::ArchType> {
+  void SetUp() override {
+    initalizeLLVM();
+    prepareElf();
+    initializeBOLT();
+  }
+
+protected:
+  void initalizeLLVM() {
+    llvm::InitializeAllTargetInfos();
+    llvm::InitializeAllTargetMCs();
+    llvm::InitializeAllAsmParsers();
+    llvm::InitializeAllDisassemblers();
+    llvm::InitializeAllTargets();
+    llvm::InitializeAllAsmPrinters();
+  }
+
+  void prepareElf() {
+    memcpy(ElfBuf, "\177ELF", 4);
+    ELF64LE::Ehdr *EHdr = reinterpret_cast<typename ELF64LE::Ehdr *>(ElfBuf);
+    EHdr->e_ident[llvm::ELF::EI_CLASS] = llvm::ELF::ELFCLASS64;
+    EHdr->e_ident[llvm::ELF::EI_DATA] = llvm::ELF::ELFDATA2LSB;
+    EHdr->e_machine = GetParam() == Triple::aarch64 ? EM_AARCH64 : EM_X86_64;
+    MemoryBufferRef Source(StringRef(ElfBuf, sizeof(ElfBuf)), "ELF");
+    ObjFile = cantFail(ObjectFile::createObjectFile(Source));
+  }
+
+  void initializeBOLT() {
+    Relocation::Arch = ObjFile->makeTriple().getArch();
+    BC = cantFail(BinaryContext::createBinaryContext(
+        ObjFile->makeTriple(), ObjFile->getFileName(), nullptr, true,
+        DWARFContext::create(*ObjFile.get()), {llvm::outs(), llvm::errs()}));
+    ASSERT_FALSE(!BC);
+  }
+
+  char ElfBuf[sizeof(typename ELF64LE::Ehdr)] = {};
+  std::unique_ptr<ObjectFile> ObjFile;
+  std::unique_ptr<BinaryContext> BC;
+};
+} // namespace
+
+#ifdef X86_AVAILABLE
+
+INSTANTIATE_TEST_SUITE_P(X86, MemoryMapsTester,
+                         ::testing::Values(Triple::x86_64));
+
+#endif
+
+#ifdef AARCH64_AVAILABLE
+
+INSTANTIATE_TEST_SUITE_P(AArch64, MemoryMapsTester,
+                         ::testing::Values(Triple::aarch64));
+
+#endif
+
+/// Check that the correct mmap size is computed when we have multiple text
+/// segment mappings.
+TEST_P(MemoryMapsTester, ParseMultipleSegments) {
+  const int Pid = 1234;
+  StringRef Filename = "BINARY";
+  opts::ReadPerfEvents = formatv(
+      "name       0 [000]     0.000000: PERF_RECORD_MMAP2 {0}/{0}: "
+      "[0xabc0000000(0x1000000) @ 0x11c0000 103:01 1573523 0]: r-xp {1}\n"
+      "name       0 [000]     0.000000: PERF_RECORD_MMAP2 {0}/{0}: "
+      "[0xabc2000000(0x8000000) @ 0x31d0000 103:01 1573523 0]: r-xp {1}\n",
+      Pid, Filename);
+
+  BC->SegmentMapInfo[0x11da000] =
+      SegmentInfo{0x11da000, 0x10da000, 0x11ca000, 0x10da000, 0x10000, true};
+  BC->SegmentMapInfo[0x31d0000] =
+      SegmentInfo{0x31d0000, 0x51ac82c, 0x31d0000, 0x3000000, 0x200000, true};
+
+  DataAggregator DA("");
+  BC->setFilename(Filename);
+  Error Err = DA.preprocessProfile(*BC);
+
+  // Ignore errors from perf2bolt when parsing memory events later on.
+  ASSERT_THAT_ERROR(std::move(Err), Succeeded());
+
+  auto &BinaryMMapInfo = DA.getBinaryMMapInfo();
+  auto El = BinaryMMapInfo.find(Pid);
+  // Check that memory mapping is present and has the expected size.
+  ASSERT_NE(El, BinaryMMapInfo.end());
+  ASSERT_EQ(El->second.Size, static_cast<uint64_t>(0xb1d0000));
+}
+
+/// Check that DataAggregator aborts when pre-processing an input binary
+/// with multiple text segments that have different base addresses.
+TEST_P(MemoryMapsTester, MultipleSegmentsMismatchedBaseAddress) {
+  const int Pid = 1234;
+  StringRef Filename = "BINARY";
+  opts::ReadPerfEvents = formatv(
+      "name       0 [000]     0.000000: PERF_RECORD_MMAP2 {0}/{0}: "
+      "[0xabc0000000(0x1000000) @ 0x11c0000 103:01 1573523 0]: r-xp {1}\n"
+      "name       0 [000]     0.000000: PERF_RECORD_MMAP2 {0}/{0}: "
+      "[0xabc2000000(0x8000000) @ 0x31d0000 103:01 1573523 0]: r-xp {1}\n",
+      Pid, Filename);
+
+  BC->SegmentMapInfo[0x11da000] =
+      SegmentInfo{0x11da000, 0x10da000, 0x11ca000, 0x10da000, 0x10000, true};
+  // Using '0x31d0fff' FileOffset which triggers a different base address
+  // for this second text segment.
+  BC->SegmentMapInfo[0x31d0000] =
+      SegmentInfo{0x31d0000, 0x51ac82c, 0x31d0fff, 0x3000000, 0x200000, true};
+
+  DataAggregator DA("");
+  BC->setFilename(Filename);
+  ASSERT_DEATH(
+      { Error Err = DA.preprocessProfile(*BC); },
+      "Base address on multiple segment mappings should match");
+}

bolt/utils/bughunter.sh

@@ -131,7 +131,7 @@ if [[ $FAIL -eq "0" ]]; then
         fi
     else
         echo "Did it pass? Type the return code [0 = pass, 1 = fail]"
-        read -n1 PASS
+        read -n1 FAIL
     fi
     if [[ $FAIL -eq "0" ]] ; then
         echo "  Warning: optimized binary passes."
@@ -205,7 +205,7 @@ while [[ "$CONTINUE" -ne "0" ]] ; do
             echo "  OPTIMIZED_BINARY failure=$FAIL"
         else
             echo "Did it pass? Type the return code [0 = pass, 1 = fail]"
-            read -n1 PASS
+            read -n1 FAIL
         fi
     else
         FAIL=1

clang/docs/SanitizerCoverage.rst

@@ -385,6 +385,20 @@ Users need to implement a single function to capture the CF table at startup:
     // the collected control flow.
   }
 
+Gated Trace Callbacks
+=====================
+
+Gate the invocation of the tracing callbacks with
+``-sanitizer-coverage-gated-trace-callbacks``.
+
+When this option is enabled, the instrumentation will not call into the
+runtime-provided callbacks for tracing, thus only incurring in a trivial
+branch without going through a function call.
+
+It is up to the runtime to toggle the value of the global variable in order to
+enable tracing.
+
+This option is only supported for trace-pc-guard and trace-cmp.
 
 Disabling instrumentation with ``__attribute__((no_sanitize("coverage")))``
 ===========================================================================

clang/lib/AST/ByteCode/Compiler.cpp

@@ -448,6 +448,10 @@ bool Compiler<Emitter>::VisitCastExpr(const CastExpr *CE) {
 
     QualType SubExprTy = SubExpr->getType();
     std::optional<PrimType> FromT = classify(SubExprTy);
+    // Casts from integer to vectors in C.
+    if (FromT && CE->getType()->isVectorType())
+      return this->emitBuiltinBitCast(CE);
+
     std::optional<PrimType> ToT = classify(CE->getType());
     if (!FromT || !ToT)
       return false;
@@ -6494,8 +6498,23 @@ bool Compiler<Emitter>::emitBuiltinBitCast(const CastExpr *E) {
   }
 
   // Get a pointer to the value-to-cast on the stack.
-  if (!this->visit(SubExpr))
-    return false;
+  // For CK_LValueToRValueBitCast, this is always an lvalue and
+  // we later assume it to be one (i.e. a PT_Ptr). However,
+  // we call this function for other utility methods where
+  // a bitcast might be useful, so convert it to a PT_Ptr in that case.
+  if (SubExpr->isGLValue()) {
+    if (!this->visit(SubExpr))
+      return false;
+  } else if (std::optional<PrimType> FromT = classify(SubExpr)) {
+    unsigned TempOffset = allocateLocalPrimitive(
+        SubExpr, *FromT, /*IsConst=*/true, /*IsExtended=*/false);
+    if (!this->visit(SubExpr))
+      return false;
+    if (!this->emitSetLocal(*FromT, TempOffset, E))
+      return false;
+    if (!this->emitGetPtrLocal(TempOffset, E))
+      return false;
+  }
 
   if (!ToT || ToT == PT_Ptr) {
     if (!this->emitBitCastPtr(E))

clang/test/AST/ByteCode/c23.c

@@ -1,5 +1,8 @@
 // RUN: %clang_cc1 -std=c23 -fexperimental-new-constant-interpreter -verify=expected,both %s
 // RUN: %clang_cc1 -std=c23 -verify=ref,both %s
+// RUN: %clang_cc1 -std=c23 -triple=aarch64_be-linux-gnu -fexperimental-new-constant-interpreter -verify=expected,both %s
+// RUN: %clang_cc1 -std=c23 -triple=aarch64_be-linux-gnu -verify=ref,both %s
+
 
 typedef typeof(nullptr) nullptr_t;
 
@@ -23,5 +26,26 @@ char bar() {
   return ((struct S *)buffer)->c;
 }
 
-
 static_assert((nullptr_t){} == 0);
+
+#if __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__
+#  define LITTLE_END 1
+#elif __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__
+#  define LITTLE_END 0
+#else
+#  error "huh?"
+#endif
+
+typedef unsigned char u8x4_t __attribute__((vector_size(4)));
+constexpr u8x4_t arg1 = (u8x4_t)0xCAFEBABE; // okay
+#if LITTLE_END
+static_assert(arg1[0] == 190);
+static_assert(arg1[1] == 186);
+static_assert(arg1[2] == 254);
+static_assert(arg1[3] == 202);
+#else
+static_assert(arg1[0] == 202);
+static_assert(arg1[1] == 254);
+static_assert(arg1[2] == 186);
+static_assert(arg1[3] == 190);
+#endif