c# - Is there a race condition in this common pattern used to prevent NullReferenceException?

Question

I asked this question and got this interesting (and a little disconcerting) answer.

Daniel states in his answer (unless I'm reading it incorrectly) that the ECMA-335 CLI specification could allow a compiler to generate code that throws a NullReferenceException from the following DoCallback method.

class MyClass {
    private Action _Callback;
    public Action Callback { 
        get { return _Callback; }
        set { _Callback = value; }
    }
    public void DoCallback() {
        Action local;
        local = Callback;
        if (local == null)
            local = new Action(() => { });
        local();
    }
}

He says that, in order to guarantee a NullReferenceException is not thrown, the volatile keyword should be used on _Callback or a lock should be used around the line local = Callback;.

Can anyone corroborate that? And, if it's true, is there a difference in behavior between Mono and .NET compilers regarding this issue?

Edit
Here is a link to the standard.

Update
I think this is the pertinent part of the spec (12.6.4):

Conforming implementations of the CLI are free to execute programs using any technology that guarantees, within a single thread of execution, that side-effects and exceptions generated by a thread are visible in the order specified by the CIL. For this purpose only volatile operations (including volatile reads) constitute visible side-effects. (Note that while only volatile operations constitute visible side-effects, volatile operations also affect the visibility of non-volatile references.) Volatile operations are specified in §12.6.7. There are no ordering guarantees relative to exceptions injected into a thread by another thread (such exceptions are sometimes called "asynchronous exceptions" (e.g., System.Threading.ThreadAbortException).

[Rationale: An optimizing compiler is free to reorder side-effects and synchronous exceptions to the extent that this reordering does not change any observable program behavior. end rationale]

[Note: An implementation of the CLI is permitted to use an optimizing compiler, for example, to convert CIL to native machine code provided the compiler maintains (within each single thread of execution) the same order of side-effects and synchronous exceptions.

So... I'm curious as to whether or not this statement allows a compiler to optimize the Callback property (which accesses a simple field) and the local variable to produce the following, which has the same behavior in a single thread of execution:

if (_Callback != null) _Callback();
else new Action(() => { })();

The 12.6.7 section on the volatile keyword seems to offer a solution for programmers wishing to avoid the optimization:

A volatile read has "acquire semantics" meaning that the read is guaranteed to occur prior to any references to memory that occur after the read instruction in the CIL instruction sequence. A volatile write has "release semantics" meaning that the write is guaranteed to happen after any memory references prior to the write instruction in the CIL instruction sequence. A conforming implementation of the CLI shall guarantee this semantics of volatile operations. This ensures that all threads will observe volatile writes performed by any other thread in the order they were performed. But a conforming implementation is not required to provide a single total ordering of volatile writes as seen from all threads of execution. An optimizing compiler that converts CIL to native code shall not remove any volatile operation, nor shall it coalesce multiple volatile operations into a single operation.

Answer 1

In CLR via C# (pp.?264–265), Jeffrey Richter discusses this specific problem, and acknowledges that it is possible for the local variable to be swapped out:

[T]his code could be optimized by the compiler to remove the local […] variable entirely. If this happens, this version of the code is identical to the [version that references the event/callback directly twice], so a NullReferenceException is still possible.

Richter suggests the use of Interlocked.CompareExchange<T> to definitively resolve this issue:

public void DoCallback() 
{
    Action local = Interlocked.CompareExchange(ref _Callback, null, null);
    if (local != null)
        local();
}

However, Richter acknowledges that Microsoft’s just-in-time (JIT) compiler does not optimize away the local variable; and, although this could, in theory, change, it almost certainly never will because it would cause too many applications to break as a result.

This question has already been asked and answered at length in “Allowed C# Compiler optimization on local variables and refetching value from memory”. Make sure to read the answer by xanatox and the “Understand the Impact of Low-Lock Techniques in Multithreaded Apps” article it cites. Since you asked specifically about Mono, you should pay attention to referenced “[Mono-dev] Memory Model?” mailing list message:

Right now we provide loose semantics close to ecma backed by the architecture you're running.