[ASAN] Fix detection of stack-use-after scope for char arrays.

[vitalybuka]

Summary

Clang inserts GetElementPtrInst so findAllocaForValue was not
able to find allocas.

PR27453

[eugenis]

This needs an IR-level test that shows that proper instrumentation appears on lifetime-after-gep allocas.

[vitalybuka]

ll test

[vitalybuka]

In D22657#492216, @eugenis wrote:

This needs an IR-level test that shows that proper instrumentation appears on lifetime-after-gep allocas.

Done.

[eugenis]

lib/Transforms/Instrumentation/AddressSanitizer.cpp

2294

We definitely should not crash if we see something we don't understand.
A dbg() message at most.

[vitalybuka]

removed assert

[eugenis]

LGTM

lib/Transforms/Instrumentation/AddressSanitizer.cpp

2292

The message should probably say something about alloca or lifetime.

[vitalybuka]

updated message

lib/Transforms/Instrumentation/AddressSanitizer.cpp

for (Value *IncValue : PN->incoming_values()) {
       // Allow self-referencing phi-nodes.
       if (IncValue == PN) continue;
       AllocaInst *IncValueAI = findAllocaForValue(IncValue);
       // AI for incoming values should exist and should all be equal.
       if (IncValueAI == nullptr || (Res != nullptr && IncValueAI != Res))
         return nullptr;
       Res = IncValueAI;
     }
+  } else if (GetElementPtrInst *EP = dyn_cast<GetElementPtrInst>(V)) {
+    Res = findAllocaForValue(EP->getPointerOperand());
+  } else {
+    // If we get here we need to handle other types of instructions which can
+    // be inserted between alloca and intrinsic.
+    assert(false);
   }
   if (Res) AllocaForValue[V] = Res;
   return Res;
 }
 
 void FunctionStackPoisoner::handleDynamicAllocaCall(AllocaInst *AI) {
   IRBuilder<> IRB(AI);