[msan] Add handleIntrinsicByApplyingToShadow; support NEON tbl/tbx (llvm#114490)

This adds a general function that handles intrinsics by applying the
intrinsic to the shadows, and applies it to the specific case of Arm
NEON TBL/TBX intrinsics.

This also updates the tests from
llvm#114462

Author: thurstond

llvm/lib/Transforms/Instrumentation/MemorySanitizer.cpp

@@ -3944,6 +3944,56 @@ struct MemorySanitizerVisitor : public InstVisitor<MemorySanitizerVisitor> {
     }
   }
 
+  /// Handle intrinsics by applying the intrinsic to the shadows.
+  ///
+  /// The trailing arguments are passed verbatim to the intrinsic, though any
+  /// uninitialized trailing arguments can also taint the shadow e.g., for an
+  /// intrinsic with one trailing verbatim argument:
+  ///     out = intrinsic(var1, var2, opType)
+  /// we compute:
+  ///     shadow[out] =
+  ///         intrinsic(shadow[var1], shadow[var2], opType) | shadow[opType]
+  ///
+  /// For example, this can be applied to the Arm NEON vector table intrinsics
+  /// (tbl{1,2,3,4}).
+  ///
+  /// The origin is approximated using setOriginForNaryOp.
+  void handleIntrinsicByApplyingToShadow(IntrinsicInst &I,
+                                         unsigned int trailingVerbatimArgs) {
+    IRBuilder<> IRB(&I);
+
+    assert(trailingVerbatimArgs < I.arg_size());
+
+    SmallVector<Value *, 8> ShadowArgs;
+    // Don't use getNumOperands() because it includes the callee
+    for (unsigned int i = 0; i < I.arg_size() - trailingVerbatimArgs; i++) {
+      Value *Shadow = getShadow(&I, i);
+      ShadowArgs.push_back(Shadow);
+    }
+
+    for (unsigned int i = I.arg_size() - trailingVerbatimArgs; i < I.arg_size();
+         i++) {
+      Value *Arg = I.getArgOperand(i);
+      ShadowArgs.push_back(Arg);
+    }
+
+    CallInst *CI =
+        IRB.CreateIntrinsic(I.getType(), I.getIntrinsicID(), ShadowArgs);
+    Value *CombinedShadow = CI;
+
+    // Combine the computed shadow with the shadow of trailing args
+    for (unsigned int i = I.arg_size() - trailingVerbatimArgs; i < I.arg_size();
+         i++) {
+      Value *Shadow =
+          CreateShadowCast(IRB, getShadow(&I, i), CombinedShadow->getType());
+      CombinedShadow = IRB.CreateOr(Shadow, CombinedShadow, "_msprop");
+    }
+
+    setShadow(&I, CombinedShadow);
+
+    setOriginForNaryOp(I);
+  }
+
   void visitIntrinsicInst(IntrinsicInst &I) {
     switch (I.getIntrinsicID()) {
     case Intrinsic::uadd_with_overflow:
@@ -4319,6 +4369,28 @@ struct MemorySanitizerVisitor : public InstVisitor<MemorySanitizerVisitor> {
       break;
     }
 
+    // Arm NEON vector table intrinsics have the source/table register(s) as
+    // arguments, followed by the index register. They return the output.
+    //
+    // 'TBL writes a zero if an index is out-of-range, while TBX leaves the
+    //  original value unchanged in the destination register.'
+    // Conveniently, zero denotes a clean shadow, which means out-of-range
+    // indices for TBL will initialize the user data with zero and also clean
+    // the shadow. (For TBX, neither the user data nor the shadow will be
+    // updated, which is also correct.)
+    case Intrinsic::aarch64_neon_tbl1:
+    case Intrinsic::aarch64_neon_tbl2:
+    case Intrinsic::aarch64_neon_tbl3:
+    case Intrinsic::aarch64_neon_tbl4:
+    case Intrinsic::aarch64_neon_tbx1:
+    case Intrinsic::aarch64_neon_tbx2:
+    case Intrinsic::aarch64_neon_tbx3:
+    case Intrinsic::aarch64_neon_tbx4: {
+      // The last trailing argument (index register) should be handled verbatim
+      handleIntrinsicByApplyingToShadow(I, 1);
+      break;
+    }
+
     default:
       if (!handleUnknownIntrinsic(I))
         visitInstruction(I);