================
@@ -0,0 +1,482 @@
+#include "AMDGPU.h"
+#include "AMDGPUTargetMachine.h"
+#include "Utils/AMDGPUBaseInfo.h"
+#include "Utils/AMDGPUMemoryUtils.h"
+#include "llvm/ADT/BitVector.h"
+#include "llvm/ADT/DenseMap.h"
+#include "llvm/ADT/DenseSet.h"
+#include "llvm/ADT/STLExtras.h"
+#include "llvm/ADT/SetOperations.h"
+#include "llvm/ADT/StringExtras.h"
+#include "llvm/ADT/StringMap.h"
+#include "llvm/ADT/StringRef.h"
+#include "llvm/Analysis/CallGraph.h"
+#include "llvm/Analysis/DomTreeUpdater.h"
+#include "llvm/CodeGen/TargetPassConfig.h"
+#include "llvm/IR/Constants.h"
+#include "llvm/IR/DerivedTypes.h"
+#include "llvm/IR/IRBuilder.h"
+#include "llvm/IR/InlineAsm.h"
+#include "llvm/IR/Instructions.h"
+#include "llvm/IR/IntrinsicsAMDGPU.h"
+#include "llvm/IR/MDBuilder.h"
+#include "llvm/IR/ReplaceConstant.h"
+#include "llvm/InitializePasses.h"
+#include "llvm/Pass.h"
+#include "llvm/Support/CommandLine.h"
+#include "llvm/Support/Debug.h"
+#include "llvm/Support/Format.h"
+#include "llvm/Support/OptimizedStructLayout.h"
+#include "llvm/Support/raw_ostream.h"
+#include "llvm/Transforms/Utils/BasicBlockUtils.h"
+#include "llvm/Transforms/Utils/ModuleUtils.h"
+
+#include <iostream>
+#define DEBUG_TYPE "amdgpu-asan-instrument-lds"
+
+using namespace llvm;
+using DomTreeCallback = function_ref<DominatorTree *(Function &F)>;
+
+namespace {
+// TODO: Just for testing purpose. Will be removed.
+cl::opt<bool> ReplaceLDSAndInstrument(
+ "amdgpu-replace-lds-and-instrument",
+ cl::desc("Replace LDS accesses with malloc and don't do asan instrumentation."),
+ cl::init(true), cl::Hidden);
+
+const char kAMDGPUBallotName[] = "llvm.amdgcn.ballot.i64";
+const char kAMDGPUUnreachableName[] = "llvm.amdgcn.unreachable";
+static const uint64_t kSmallX86_64ShadowOffsetBase = 0x7FFFFFFF;
+static const uint64_t kSmallX86_64ShadowOffsetAlignMask = ~0xFFFULL;
+const bool Recover = true;
+const uint32_t AsanMappingScale = 3;
+const uint32_t AsanMappingOffset =
+ (kSmallX86_64ShadowOffsetBase &
+ (kSmallX86_64ShadowOffsetAlignMask << AsanMappingScale));
+
+class AMDGPUAsanInstrumentLDS : public ModulePass {
+
+public:
+ static char ID;
+ AMDGPUAsanInstrumentLDS() : ModulePass(ID) {}
+
+ bool runOnModule(Module &M) override;
+
+ void getAnalysisUsage(AnalysisUsage &AU) const override {
+ AU.addPreserved<DominatorTreeWrapperPass>();
+ AU.addRequiredID(AMDGPULowerModuleLDSLegacyPassID);
+ }
+};
+} // namespace
+
+INITIALIZE_PASS(AMDGPUAsanInstrumentLDS, "amdgpu-asan-instrument-lds",
+ "AMDGPU AddressSanitizer instrument LDS", false, false)
+
+char AMDGPUAsanInstrumentLDS::ID = 0;
+
+static uint64_t getRedzoneSizeForGlobal(uint64_t SizeInBytes) {
+ constexpr uint64_t kMaxRZ = 1 << 18;
+ // TODO: get scale from asan-mapping-scale
+ const int MappingScale = AsanMappingScale;
+ const uint64_t MinRZ = std::max(32U, 1U << MappingScale);
+ ;
+
+ uint64_t RZ = 0;
+ if (SizeInBytes <= MinRZ / 2) {
+ // Reduce redzone size for small size objects, e.g. int, char[1]. MinRZ is
+ // at least 32 bytes, optimize when SizeInBytes is less than or equal to
+ // half of MinRZ.
+ RZ = MinRZ - SizeInBytes;
+ } else {
+ // Calculate RZ, where MinRZ <= RZ <= MaxRZ, and RZ ~ 1/4 * SizeInBytes.
+ RZ = std::clamp((SizeInBytes / MinRZ / 4) * MinRZ, MinRZ, kMaxRZ);
+
+ // Round up to multiple of MinRZ.
+ if (SizeInBytes % MinRZ)
+ RZ += MinRZ - (SizeInBytes % MinRZ);
+ }
+
+ assert((RZ + SizeInBytes) % MinRZ == 0);
+
+ return RZ;
+}
+
+static Instruction *genAMDGPUReportBlock(Module &M, IRBuilder<> &IRB,
+ Value *Cond, bool Recover) {
+ Value *ReportCond = Cond;
+ if (!Recover) {
+ auto Ballot = M.getOrInsertFunction(kAMDGPUBallotName, IRB.getInt64Ty(),
+ IRB.getInt1Ty());
+ ReportCond = IRB.CreateIsNotNull(IRB.CreateCall(Ballot, {Cond}));
+ }
+
+ auto *Trm = SplitBlockAndInsertIfThen(
+ ReportCond, &*IRB.GetInsertPoint(), false,
+ MDBuilder(M.getContext()).createBranchWeights(1, 100000));
+ Trm->getParent()->setName("asan.report");
+
+ if (Recover)
+ return Trm;
+
+ Trm = SplitBlockAndInsertIfThen(Cond, Trm, false);
+ IRB.SetInsertPoint(Trm);
+ return IRB.CreateCall(
+ M.getOrInsertFunction(kAMDGPUUnreachableName, IRB.getVoidTy()), {});
+}
+
+static Value *createSlowPathCmp(Module &M, IRBuilder<> &IRB, Value *AddrLong,
+ Value *ShadowValue, uint32_t TypeStoreSize) {
+
+ unsigned int LongSize = M.getDataLayout().getPointerSizeInBits();
+ IntegerType *IntptrTy = Type::getIntNTy(M.getContext(), LongSize);
+ size_t Granularity = static_cast<size_t>(1) << AsanMappingScale;
+ // Addr & (Granularity - 1)
+ Value *LastAccessedByte =
+ IRB.CreateAnd(AddrLong, ConstantInt::get(IntptrTy, Granularity - 1));
----------------
arsenm wrote:
Can this use llvm.ptrmask?
https://github.com/llvm/llvm-project/pull/83287