[DependenceAnalysis] Check if result of getConstantPart is null

[bcahoon]

Summary

A seg-fault occurs due to a reference of a null pointer, which is the value returned by getConstantPart. This function returns null if the constant part is not found. The code that calls this function needs to check for the null return value.

[sanjoy]

Hi Brendan,

I'm not familiar with this code, but this change looks innocent enough that I LGTM'ed it.

However, I have a design nit: it looks like getConstantPart should be defined to take a const SCEV * and if it is a SCEVConstant, return that, else if it is SCEVMulExpr do what it does today and otherwise return nullptr -- it looks like that will fit in better with how getConstantPart is used. Moreover, a SCEVConstantExpr is really a degenerate SCEVMulExpr :) Do you mind doing that refactoring in a later change?

[bcahoon]

Thanks for the review, Sanjoy. I'll refactor the getConstantPart function in a subsequent change as suggested.

llvm/trunk/lib/Analysis/DependenceAnalysis.cpp

while (RunningGCD != 1 && isa<SCEVAddRecExpr>(Inner)) {
         ; // SrcCoeff == Coeff
       else {
         if (const SCEVMulExpr *Product = dyn_cast<SCEVMulExpr>(Coeff))
           // If the coefficient is the product of a constant and other stuff,
           // we can use the constant in the GCD computation.
           Constant = getConstantPart(Product);
         else
           Constant = cast<SCEVConstant>(Coeff);
+        if (!Constant)
+          return false;
         APInt ConstCoeff = Constant->getAPInt();
         RunningGCD = APIntOps::GreatestCommonDivisor(RunningGCD, ConstCoeff.abs());
       }
       Inner = AddRec->getStart();
     }
     Inner = Dst;
     while (RunningGCD != 1 && isa<SCEVAddRecExpr>(Inner)) {
       AddRec = cast<SCEVAddRecExpr>(Inner);
       const SCEV *Coeff = AddRec->getStepRecurrence(*SE);
       if (CurLoop == AddRec->getLoop())
         DstCoeff = Coeff;
       else {
         if (const SCEVMulExpr *Product = dyn_cast<SCEVMulExpr>(Coeff))
           // If the coefficient is the product of a constant and other stuff,
           // we can use the constant in the GCD computation.
           Constant = getConstantPart(Product);
         else
           Constant = cast<SCEVConstant>(Coeff);
+        if (!Constant)
+          return false;
         APInt ConstCoeff = Constant->getAPInt();
         RunningGCD = APIntOps::GreatestCommonDivisor(RunningGCD, ConstCoeff.abs());
       }
       Inner = AddRec->getStart();
     }
     Delta = SE->getMinusSCEV(SrcCoeff, DstCoeff);
     if (const SCEVMulExpr *Product = dyn_cast<SCEVMulExpr>(Delta))
       // If the coefficient is the product of a constant and other stuff,
       // we can use the constant in the GCD computation.
       Constant = getConstantPart(Product);
     else if (isa<SCEVConstant>(Delta))
       Constant = cast<SCEVConstant>(Delta);
     else {
       // The difference of the two coefficients might not be a product
       // or constant, in which case we give up on this direction.
       continue;
     }
+    if (!Constant)
+      continue;
     APInt ConstCoeff = Constant->getAPInt();
     RunningGCD = APIntOps::GreatestCommonDivisor(RunningGCD, ConstCoeff.abs());
     DEBUG(dbgs() << "\tRunningGCD = " << RunningGCD << "\n");
     if (RunningGCD != 0) {
       Remainder = ConstDelta.srem(RunningGCD);
       DEBUG(dbgs() << "\tRemainder = " << Remainder << "\n");
       if (Remainder != 0) {
         unsigned Level = mapSrcLoop(CurLoop);

llvm/trunk/test/Analysis/DependenceAnalysis/MIVCheckConst.ll

+; RUN: opt < %s -analyze -basicaa -da
+
+; Test that the dependence analysis pass does seg-fault due to a null pointer
+; dereference. The code in gcdMIVTest requires a null check for the result of
+; getConstantPart.
+
+target datalayout = "e-m:e-p:32:32-i1:32-i64:64-a:0-v32:32-n16:32"
+
+%0 = type { i32 }
+%1 = type { [2 x [512 x %0]], [512 x %0], %2, [144 x i8], %9, %10, %11, %12, %17, [12 x i8], %18, %19, %21, [128 x i8] }
+%2 = type { [64 x i16], [64 x i16], [64 x %0], [64 x %0], [128 x %0], [128 x %0], [256 x %0], [256 x %0], [32 x %0], [32 x %0], [64 x i16], [64 x i16], [64 x i16], [64 x i16], [64 x i16], %3, %4, %5, [32 x i32], [32 x i32], [32 x i32], [32 x i32], [32 x i32], [32 x i32], %6, %7, [32 x i32], [32 x i32], [32 x i32], [64 x i16], %8, [8 x i64], [4 x i64], [2 x i64], [256 x i8], [256 x i32], [64 x i16], [64 x i16] }
+%3 = type { [64 x i32], [32 x i32], [32 x i32], [32 x i32], [32 x i32], [32 x i32], [32 x i32], [32 x i32], [32 x i32], [32 x i32], [32 x i32], [32 x i32], [32 x i32], [32 x i32], [32 x i32], [64 x i32], [64 x i32], [64 x i32], [64 x i32], [64 x i32], [64 x i32], [64 x i32], [64 x i32], [64 x i32], [64 x i16], [64 x i16], [64 x i16], [64 x i16], [64 x i16], [64 x i16], [64 x i16], [64 x i16], [64 x i16], [64 x i16], [64 x i16], [64 x i16], [32 x %0], [32 x %0], [128 x i8] }
+%4 = type { [64 x i16], [64 x i16], [64 x i16], [64 x i16], [64 x i16], [64 x i16], [64 x i16], [64 x i16] }
+%5 = type { [128 x i8], [256 x i8], [256 x i8] }
+%6 = type { [64 x i32], [128 x i16], [64 x i16], [64 x i16], [64 x i16] }
+%7 = type { [192 x %0], [192 x %0], [384 x %0], [1984 x %0] }
+%8 = type { [128 x i8], [128 x i8], [64 x i16], [64 x i16], [64 x i16], [64 x i16], [64 x i16], [64 x i16] }
+%9 = type { [32 x %0], [32 x %0], [64 x i32], [64 x i32], [64 x i32], [64 x i32] }
+%10 = type { [1536 x %0], [2048 x %0], [512 x i32], [256 x i32], [32 x %0], [64 x i32], [128 x i8], [512 x i32], [1024 x %0] }
+%11 = type { [512 x i32], [512 x i32], [1024 x %0], [512 x i32], [512 x %0] }
+%12 = type { %13, [2048 x %0], [2048 x %0], [256 x i32], [1024 x i32], %14, [512 x %0], [256 x i32], %15, [4 x [256 x %0]], [4 x [256 x %0]], [256 x i32], [8 x [32 x i32]], [8 x [32 x %0]], [384 x %0], [256 x i32], %16 }
+%13 = type { [2048 x %0] }
+%14 = type { [1024 x %0], [1024 x %0] }
+%15 = type { [256 x %0], [256 x %0] }
+%16 = type { [128 x %0], [128 x %0] }
+%17 = type { [32 x %0], [32 x i32], [32 x i32], [32 x i32], [32 x i32], [2 x [8 x [32 x %0]]], [512 x %0], [512 x %0], [58 x i16] }
+%18 = type { [512 x i8] }
+%19 = type { [2048 x %0], [2560 x i16], %20, [512 x i32], [256 x i32], [512 x i8] }
+%20 = type { [768 x i32] }
+%21 = type { [416 x i32] }
+
+define void @test(%1* %A) #0 align 2 {
+entry:
+  %v1 = load i32, i32* undef, align 4
+  br label %bb13
+
+bb13:
+  %v2 = phi i32 [ undef, %entry ], [ %v39, %bb38 ]
+  br i1 undef, label %bb15, label %bb38
+
+bb15:
+  %v3 = mul nsw i32 %v2, undef
+  br label %bb17
+
+bb17:
+  br i1 undef, label %bb21, label %bb37
+
+bb21:
+  %v22 = add nsw i32 undef, 1
+  %v23 = add i32 %v22, %v3
+  %v24 = mul nsw i32 %v23, %v1
+  %v25 = getelementptr inbounds %1, %1* %A, i32 0, i32 7, i32 1, i32 %v24
+  %v26 = bitcast %0* %v25 to <32 x i32>*
+  %v27 = load <32 x i32>, <32 x i32>* %v26, align 256
+  %v28 = add i32 undef, %v3
+  %v29 = mul nsw i32 %v28, 32
+  %v30 = getelementptr inbounds %1, %1* %A, i32 0, i32 7, i32 14, i32 %v29
+  %v31 = bitcast %0* %v30 to <32 x i32>*
+  %v32 = load <32 x i32>, <32 x i32>* %v31, align 128
+  br i1 undef, label %bb21, label %bb37
+
+bb37:
+  br i1 undef, label %bb17, label %bb38
+
+bb38:
+  %v39 = add nsw i32 %v2, 1
+  br label %bb13
+
+bb40:
+  ret void
+}
+
+attributes #0 = { "less-precise-fpmad"="false" "no-frame-pointer-elim"="true" "no-frame-pointer-elim-non-leaf" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }