[AArch64] Fix movk parsing with an .equ operand (#124428)
[llvm-project.git] / llvm / lib / Transforms / Instrumentation / BoundsChecking.cpp
blob609678f9979c635d378c973cd510334a48bb907f
1 //===- BoundsChecking.cpp - Instrumentation for run-time bounds checking --===//
2 //
3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4 // See https://llvm.org/LICENSE.txt for license information.
5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6 //
7 //===----------------------------------------------------------------------===//
9 #include "llvm/Transforms/Instrumentation/BoundsChecking.h"
10 #include "llvm/ADT/Statistic.h"
11 #include "llvm/ADT/StringRef.h"
12 #include "llvm/ADT/Twine.h"
13 #include "llvm/Analysis/MemoryBuiltins.h"
14 #include "llvm/Analysis/ScalarEvolution.h"
15 #include "llvm/Analysis/TargetFolder.h"
16 #include "llvm/Analysis/TargetLibraryInfo.h"
17 #include "llvm/IR/BasicBlock.h"
18 #include "llvm/IR/Constants.h"
19 #include "llvm/IR/DataLayout.h"
20 #include "llvm/IR/Function.h"
21 #include "llvm/IR/IRBuilder.h"
22 #include "llvm/IR/InstIterator.h"
23 #include "llvm/IR/Instruction.h"
24 #include "llvm/IR/Instructions.h"
25 #include "llvm/IR/Intrinsics.h"
26 #include "llvm/IR/Value.h"
27 #include "llvm/Support/Casting.h"
28 #include "llvm/Support/CommandLine.h"
29 #include "llvm/Support/Debug.h"
30 #include "llvm/Support/raw_ostream.h"
31 #include <utility>
33 using namespace llvm;
35 #define DEBUG_TYPE "bounds-checking"
37 static cl::opt<bool> SingleTrapBB("bounds-checking-single-trap",
38 cl::desc("Use one trap block per function"));
40 STATISTIC(ChecksAdded, "Bounds checks added");
41 STATISTIC(ChecksSkipped, "Bounds checks skipped");
42 STATISTIC(ChecksUnable, "Bounds checks unable to add");
44 class BuilderTy : public IRBuilder<TargetFolder> {
45 public:
46 BuilderTy(BasicBlock *TheBB, BasicBlock::iterator IP, TargetFolder Folder)
47 : IRBuilder<TargetFolder>(TheBB, IP, Folder) {
48 SetNoSanitizeMetadata();
52 /// Gets the conditions under which memory accessing instructions will overflow.
53 ///
54 /// \p Ptr is the pointer that will be read/written, and \p InstVal is either
55 /// the result from the load or the value being stored. It is used to determine
56 /// the size of memory block that is touched.
57 ///
58 /// Returns the condition under which the access will overflow.
59 static Value *getBoundsCheckCond(Value *Ptr, Value *InstVal,
60 const DataLayout &DL, TargetLibraryInfo &TLI,
61 ObjectSizeOffsetEvaluator &ObjSizeEval,
62 BuilderTy &IRB, ScalarEvolution &SE) {
63 TypeSize NeededSize = DL.getTypeStoreSize(InstVal->getType());
64 LLVM_DEBUG(dbgs() << "Instrument " << *Ptr << " for " << Twine(NeededSize)
65 << " bytes\n");
67 SizeOffsetValue SizeOffset = ObjSizeEval.compute(Ptr);
69 if (!SizeOffset.bothKnown()) {
70 ++ChecksUnable;
71 return nullptr;
74 Value *Size = SizeOffset.Size;
75 Value *Offset = SizeOffset.Offset;
76 ConstantInt *SizeCI = dyn_cast<ConstantInt>(Size);
78 Type *IndexTy = DL.getIndexType(Ptr->getType());
79 Value *NeededSizeVal = IRB.CreateTypeSize(IndexTy, NeededSize);
81 auto SizeRange = SE.getUnsignedRange(SE.getSCEV(Size));
82 auto OffsetRange = SE.getUnsignedRange(SE.getSCEV(Offset));
83 auto NeededSizeRange = SE.getUnsignedRange(SE.getSCEV(NeededSizeVal));
85 // three checks are required to ensure safety:
86 // . Offset >= 0 (since the offset is given from the base ptr)
87 // . Size >= Offset (unsigned)
88 // . Size - Offset >= NeededSize (unsigned)
90 // optimization: if Size >= 0 (signed), skip 1st check
91 // FIXME: add NSW/NUW here? -- we dont care if the subtraction overflows
92 Value *ObjSize = IRB.CreateSub(Size, Offset);
93 Value *Cmp2 = SizeRange.getUnsignedMin().uge(OffsetRange.getUnsignedMax())
94 ? ConstantInt::getFalse(Ptr->getContext())
95 : IRB.CreateICmpULT(Size, Offset);
96 Value *Cmp3 = SizeRange.sub(OffsetRange)
97 .getUnsignedMin()
98 .uge(NeededSizeRange.getUnsignedMax())
99 ? ConstantInt::getFalse(Ptr->getContext())
100 : IRB.CreateICmpULT(ObjSize, NeededSizeVal);
101 Value *Or = IRB.CreateOr(Cmp2, Cmp3);
102 if ((!SizeCI || SizeCI->getValue().slt(0)) &&
103 !SizeRange.getSignedMin().isNonNegative()) {
104 Value *Cmp1 = IRB.CreateICmpSLT(Offset, ConstantInt::get(IndexTy, 0));
105 Or = IRB.CreateOr(Cmp1, Or);
108 return Or;
111 static CallInst *InsertTrap(BuilderTy &IRB, bool DebugTrapBB) {
112 if (!DebugTrapBB)
113 return IRB.CreateIntrinsic(Intrinsic::trap, {}, {});
114 // FIXME: Ideally we would use the SanitizerHandler::OutOfBounds constant.
115 return IRB.CreateIntrinsic(
116 Intrinsic::ubsantrap, {},
117 ConstantInt::get(IRB.getInt8Ty(),
118 IRB.GetInsertBlock()->getParent()->size()));
121 static CallInst *InsertCall(BuilderTy &IRB, bool MayReturn, StringRef Name) {
122 Function *Fn = IRB.GetInsertBlock()->getParent();
123 LLVMContext &Ctx = Fn->getContext();
124 llvm::AttrBuilder B(Ctx);
125 B.addAttribute(llvm::Attribute::NoUnwind);
126 if (!MayReturn)
127 B.addAttribute(llvm::Attribute::NoReturn);
128 FunctionCallee Callee = Fn->getParent()->getOrInsertFunction(
129 Name,
130 llvm::AttributeList::get(Ctx, llvm::AttributeList::FunctionIndex, B),
131 Type::getVoidTy(Ctx));
132 return IRB.CreateCall(Callee);
135 /// Adds run-time bounds checks to memory accessing instructions.
137 /// \p Or is the condition that should guard the trap.
139 /// \p GetTrapBB is a callable that returns the trap BB to use on failure.
140 template <typename GetTrapBBT>
141 static void insertBoundsCheck(Value *Or, BuilderTy &IRB, GetTrapBBT GetTrapBB) {
142 // check if the comparison is always false
143 ConstantInt *C = dyn_cast_or_null<ConstantInt>(Or);
144 if (C) {
145 ++ChecksSkipped;
146 // If non-zero, nothing to do.
147 if (!C->getZExtValue())
148 return;
150 ++ChecksAdded;
152 BasicBlock::iterator SplitI = IRB.GetInsertPoint();
153 BasicBlock *OldBB = SplitI->getParent();
154 BasicBlock *Cont = OldBB->splitBasicBlock(SplitI);
155 OldBB->getTerminator()->eraseFromParent();
157 BasicBlock *TrapBB = GetTrapBB(IRB, Cont);
159 if (C) {
160 // If we have a constant zero, unconditionally branch.
161 // FIXME: We should really handle this differently to bypass the splitting
162 // the block.
163 BranchInst::Create(TrapBB, OldBB);
164 return;
167 // Create the conditional branch.
168 BranchInst::Create(TrapBB, Cont, Or, OldBB);
171 static std::string
172 getRuntimeCallName(const BoundsCheckingPass::Options::Runtime &Opts) {
173 std::string Name = "__ubsan_handle_local_out_of_bounds";
174 if (Opts.MinRuntime)
175 Name += "_minimal";
176 if (!Opts.MayReturn)
177 Name += "_abort";
178 return Name;
181 static bool addBoundsChecking(Function &F, TargetLibraryInfo &TLI,
182 ScalarEvolution &SE,
183 const BoundsCheckingPass::Options &Opts) {
184 if (F.hasFnAttribute(Attribute::NoSanitizeBounds))
185 return false;
187 const DataLayout &DL = F.getDataLayout();
188 ObjectSizeOpts EvalOpts;
189 EvalOpts.RoundToAlign = true;
190 EvalOpts.EvalMode = ObjectSizeOpts::Mode::ExactUnderlyingSizeAndOffset;
191 ObjectSizeOffsetEvaluator ObjSizeEval(DL, &TLI, F.getContext(), EvalOpts);
193 // check HANDLE_MEMORY_INST in include/llvm/Instruction.def for memory
194 // touching instructions
195 SmallVector<std::pair<Instruction *, Value *>, 4> TrapInfo;
196 for (Instruction &I : instructions(F)) {
197 Value *Or = nullptr;
198 BuilderTy IRB(I.getParent(), BasicBlock::iterator(&I), TargetFolder(DL));
199 if (LoadInst *LI = dyn_cast<LoadInst>(&I)) {
200 if (!LI->isVolatile())
201 Or = getBoundsCheckCond(LI->getPointerOperand(), LI, DL, TLI,
202 ObjSizeEval, IRB, SE);
203 } else if (StoreInst *SI = dyn_cast<StoreInst>(&I)) {
204 if (!SI->isVolatile())
205 Or = getBoundsCheckCond(SI->getPointerOperand(), SI->getValueOperand(),
206 DL, TLI, ObjSizeEval, IRB, SE);
207 } else if (AtomicCmpXchgInst *AI = dyn_cast<AtomicCmpXchgInst>(&I)) {
208 if (!AI->isVolatile())
209 Or =
210 getBoundsCheckCond(AI->getPointerOperand(), AI->getCompareOperand(),
211 DL, TLI, ObjSizeEval, IRB, SE);
212 } else if (AtomicRMWInst *AI = dyn_cast<AtomicRMWInst>(&I)) {
213 if (!AI->isVolatile())
214 Or = getBoundsCheckCond(AI->getPointerOperand(), AI->getValOperand(),
215 DL, TLI, ObjSizeEval, IRB, SE);
217 if (Or) {
218 if (Opts.GuardKind) {
219 llvm::Value *Allow = IRB.CreateIntrinsic(
220 IRB.getInt1Ty(), Intrinsic::allow_ubsan_check,
221 {llvm::ConstantInt::getSigned(IRB.getInt8Ty(), *Opts.GuardKind)});
222 Or = IRB.CreateAnd(Or, Allow);
224 TrapInfo.push_back(std::make_pair(&I, Or));
228 std::string Name;
229 if (Opts.Rt)
230 Name = getRuntimeCallName(*Opts.Rt);
232 // Create a trapping basic block on demand using a callback. Depending on
233 // flags, this will either create a single block for the entire function or
234 // will create a fresh block every time it is called.
235 BasicBlock *ReuseTrapBB = nullptr;
236 auto GetTrapBB = [&ReuseTrapBB, &Opts, &Name](BuilderTy &IRB,
237 BasicBlock *Cont) {
238 Function *Fn = IRB.GetInsertBlock()->getParent();
239 auto DebugLoc = IRB.getCurrentDebugLocation();
240 IRBuilder<>::InsertPointGuard Guard(IRB);
242 // Create a trapping basic block on demand using a callback. Depending on
243 // flags, this will either create a single block for the entire function or
244 // will create a fresh block every time it is called.
245 if (ReuseTrapBB)
246 return ReuseTrapBB;
248 BasicBlock *TrapBB = BasicBlock::Create(Fn->getContext(), "trap", Fn);
249 IRB.SetInsertPoint(TrapBB);
251 bool DebugTrapBB = !Opts.Merge;
252 CallInst *TrapCall = Opts.Rt ? InsertCall(IRB, Opts.Rt->MayReturn, Name)
253 : InsertTrap(IRB, DebugTrapBB);
254 if (DebugTrapBB)
255 TrapCall->addFnAttr(llvm::Attribute::NoMerge);
257 TrapCall->setDoesNotThrow();
258 TrapCall->setDebugLoc(DebugLoc);
260 bool MayReturn = Opts.Rt && Opts.Rt->MayReturn;
261 if (MayReturn) {
262 IRB.CreateBr(Cont);
263 } else {
264 TrapCall->setDoesNotReturn();
265 IRB.CreateUnreachable();
268 if (!MayReturn && SingleTrapBB && !DebugTrapBB)
269 ReuseTrapBB = TrapBB;
271 return TrapBB;
274 for (const auto &Entry : TrapInfo) {
275 Instruction *Inst = Entry.first;
276 BuilderTy IRB(Inst->getParent(), BasicBlock::iterator(Inst), TargetFolder(DL));
277 insertBoundsCheck(Entry.second, IRB, GetTrapBB);
280 return !TrapInfo.empty();
283 PreservedAnalyses BoundsCheckingPass::run(Function &F, FunctionAnalysisManager &AM) {
284 auto &TLI = AM.getResult<TargetLibraryAnalysis>(F);
285 auto &SE = AM.getResult<ScalarEvolutionAnalysis>(F);
287 if (!addBoundsChecking(F, TLI, SE, Opts))
288 return PreservedAnalyses::all();
290 return PreservedAnalyses::none();
293 void BoundsCheckingPass::printPipeline(
294 raw_ostream &OS, function_ref<StringRef(StringRef)> MapClassName2PassName) {
295 static_cast<PassInfoMixin<BoundsCheckingPass> *>(this)->printPipeline(
296 OS, MapClassName2PassName);
297 OS << "<";
298 if (Opts.Rt) {
299 if (Opts.Rt->MinRuntime)
300 OS << "min-";
301 OS << "rt";
302 if (!Opts.Rt->MayReturn)
303 OS << "-abort";
304 } else {
305 OS << "trap";
307 if (Opts.Merge)
308 OS << ";merge";
309 if (Opts.GuardKind)
310 OS << ";guard=" << static_cast<int>(*Opts.GuardKind);
311 OS << ">";