[PM] Teach SCEV to invalidate itself when its dependencies become

invalid.

This fixes use-after-free bugs that will arise with any interesting use
of SCEV.

I've added a dedicated test that works diligently to trigger these kinds
of bugs in the new pass manager and also checks for them explicitly as
well as triggering ASan failures when things go squirly.

llvm-svn: 291426

1 files changed, 70 insertions, 0 deletions

diff --git a/llvm/test/Analysis/ScalarEvolution/invalidation.ll b/llvm/test/Analysis/ScalarEvolution/invalidation.ll
new file mode 100644
index 00000000000..1fcaddb525e
--- /dev/null
+++ b/llvm/test/Analysis/ScalarEvolution/invalidation.ll
@@ -0,0 +1,70 @@
+; Test that SCEV gets invalidated when one of its dependencies is invalidated.
+;
+; Each of the RUNs checks that the pass manager runs SCEV, then invalidates it
+; due to a dependency being invalidated, and then re-urns it. This will
+; directly fail and indicates a failure that would occur later if we ddidn't
+; invalidate SCEV in this way.
+
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-linux-gnu"
+
+; RUN: opt < %s -passes='require<scalar-evolution>,invalidate<assumptions>,print<scalar-evolution>' \
+; RUN:     -debug-pass-manager -disable-output 2>&1 \
+; RUN:     | FileCheck %s -check-prefixes=CHECK,CHECK-AC-INVALIDATE
+;
+; CHECK-AC-INVALIDATE: Running pass: RequireAnalysisPass
+; CHECK-AC-INVALIDATE: Running analysis: ScalarEvolutionAnalysis
+; CHECK-AC-INVALIDATE: Running analysis: AssumptionAnalysis
+; CHECK-AC-INVALIDATE: Running pass: InvalidateAnalysisPass
+; CHECK-AC-INVALIDATE: Invalidating analysis: AssumptionAnalysis
+; CHECK-AC-INVALIDATE: Running pass: ScalarEvolutionPrinterPass
+; CHECK-AC-INVALIDATE: Running analysis: ScalarEvolutionAnalysis
+; CHECK-AC-INVALIDATE: Running analysis: AssumptionAnalysis
+
+; RUN: opt < %s -passes='require<scalar-evolution>,invalidate<domtree>,print<scalar-evolution>' \
+; RUN:     -debug-pass-manager -disable-output 2>&1 \
+; RUN:     | FileCheck %s -check-prefixes=CHECK,CHECK-DT-INVALIDATE
+;
+; CHECK-DT-INVALIDATE: Running pass: RequireAnalysisPass
+; CHECK-DT-INVALIDATE: Running analysis: ScalarEvolutionAnalysis
+; CHECK-DT-INVALIDATE: Running analysis: DominatorTreeAnalysis
+; CHECK-DT-INVALIDATE: Running pass: InvalidateAnalysisPass
+; CHECK-DT-INVALIDATE: Invalidating analysis: DominatorTreeAnalysis
+; CHECK-DT-INVALIDATE: Running pass: ScalarEvolutionPrinterPass
+; CHECK-DT-INVALIDATE: Running analysis: ScalarEvolutionAnalysis
+; CHECK-DT-INVALIDATE: Running analysis: DominatorTreeAnalysis
+
+; RUN: opt < %s -passes='require<scalar-evolution>,invalidate<loops>,print<scalar-evolution>' \
+; RUN:     -debug-pass-manager -disable-output 2>&1 \
+; RUN:     | FileCheck %s -check-prefixes=CHECK,CHECK-LI-INVALIDATE
+;
+; CHECK-LI-INVALIDATE: Running pass: RequireAnalysisPass
+; CHECK-LI-INVALIDATE: Running analysis: ScalarEvolutionAnalysis
+; CHECK-LI-INVALIDATE: Running analysis: LoopAnalysis
+; CHECK-LI-INVALIDATE: Running pass: InvalidateAnalysisPass
+; CHECK-LI-INVALIDATE: Invalidating analysis: LoopAnalysis
+; CHECK-LI-INVALIDATE: Running pass: ScalarEvolutionPrinterPass
+; CHECK-LI-INVALIDATE: Running analysis: ScalarEvolutionAnalysis
+; CHECK-LI-INVALIDATE: Running analysis: LoopAnalysis
+
+; This test isn't particularly interesting, its just enough to make sure we
+; actually do some work inside of SCEV so that if we regress here despite the
+; debug pass printing continuing to match, ASan and other tools can catch it.
+define void @test(i32 %n) {
+; CHECK-LABEL: Classifying expressions for: @test
+; CHECK: Loop %loop: backedge-taken count is 14
+; CHECK: Loop %loop: max backedge-taken count is 14
+; CHECK: Loop %loop: Predicated backedge-taken count is 14
+
+entry:
+  br label %loop
+
+loop:
+  %iv = phi i32 [ 0, %entry ], [ %iv.inc, %loop ]
+  %iv.inc = add nsw i32 %iv, 3
+  %becond = icmp ne i32 %iv.inc, 46
+  br i1 %becond, label %loop, label %leave
+
+leave:
+  ret void
+}