Skip to content

Latest commit

 

History

History
412 lines (280 loc) · 21.5 KB

span-safety.md

File metadata and controls

412 lines (280 loc) · 21.5 KB
Raw

Compile time enforcement of safety for ref-like types

Introduction

The main reason for the additional safety rules when dealing with types like Span<T> and ReadOnlySpan<T> is that such types must be confined to the execution stack.

There are two reasons why Span<T> and similar types must be a stack-only types.

  1. Span<T> is semantically a struct containing a reference and a range - (ref T data, int length). Regardless of actual implementation, writes to such struct would not be atomic. Concurrent "tearing" of such struct would lead to the possibility of length not matching the data, causing out-of-range accesses and type-safety violations, which ultimately could result in GC heap corruption in seemingly "safe" code.
  2. Some implementations of Span<T> literally contain a managed pointer in one of its fields. Managed pointers are not supported as fields of heap objects and code that manages to put a managed pointer on the GC heap typically crashes at JIT time.

All the above problems would be alleviated if instances of Span<T> are constrained to exist only on the execution stack.

An additional problem arises due to composition. It would be generally desirable to build more complex data types that would embed Span<T> and ReadOnlySpan<T> instances. Such composite types would have to be structs and would share all the hazards and requirements of Span<T>. As a result the safety rules described here should be viewed as applicable to the whole range of ref-like types.

The draft language specification is intended to ensure that values of a ref-like type occur only on the stack.

Generalized ref-like types in source code

ref-like structs are explicitly marked in the source code using ref modifier:

ref struct TwoSpans<T>
{
	// can have ref-like instance fields
	public Span<T> first;
	public Span<T> second;
} 

// error: arrays of ref-like types are not allowed. 
TwoSpans<T>[] arr = null;

Designating a struct as ref-like will allow the struct to have ref-like instance fields and will also make all the requirements of ref-like types applicable to the struct.

Metadata representation of ref-like structs

Ref-like structs will be marked with System.Runtime.CompilerServices.IsRefLikeAttribute attribute.

The attribute will be added to common base libraries such as mscorlib. In a case if the attribute is not available, compiler will generate an internal one similarly to other embedded-on-demand attributes such as IsReadOnlyAttribute.

An additional measure will be taken to prevent the use of ref-like structs in compilers not familiar with the safety rules (this includes C# compilers prior to the one in which this feature is implemented).

Having no other good alternatives that work in old compilers without servicing, an Obsolete attribute with a known string will be added to all ref-like structs. Compilers that know how to use ref-like types will ignore this particular form of Obsolete.

A typical metadata representation:

    [IsRefLike]
    [Obsolete("Types with embedded references are not supported in this version of your compiler.")]
    public struct TwoSpans<T>
    {
       // . . . .
    }

NOTE: it is not the goal to make it so that any use of ref-like types on old compilers fails 100%. That is hard to achieve and is not strictly necessary. For example there would always be a way to get around the Obsolete using dynamic code or, for example, creating an array of ref-like types through reflection.

In particular, if user wants to actually put an Obsolete or Deprecated attribute on a ref-like type, we will have no choice other than not emitting the predefined one since Obsolete attribute cannot be applied more than once..

Examples:

SpanLikeType M1(ref SpanLikeType x, Span<byte> y)
{
    // this is all valid, unconcerned with stack-referring stuff
    var local = new SpanLikeType(y);
    x = local;
    return x;
}

void Test1(ref SpanLikeType param1, Span<byte> param2)
{
    Span<byte> stackReferring1 = stackalloc byte[10];
    var stackReferring2 = new SpanLikeType(stackReferring1);

    // this is allowed
    stackReferring2 = M1(ref stackReferring2, stackReferring1);

    // this is NOT allowed
    stackReferring2 = M1(ref param1, stackReferring1);

    // this is NOT allowed
    param1 = M1(ref stackReferring2, stackReferring1);

    // this is NOT allowed
    param2 = stackReferring1.Slice(10);

    // this is allowed
    param1 = new SpanLikeType(param2);

    // this is allowed
    stackReferring2 = param1;
}

ref SpanLikeType M2(ref SpanLikeType x)
{
    return ref x;
}

ref SpanLikeType Test2(ref SpanLikeType param1, Span<byte> param2)
{
    Span<byte> stackReferring1 = stackalloc byte[10];
    var stackReferring2 = new SpanLikeType(stackReferring1);

    ref var stackReferring3 = M2(ref stackReferring2);

    // this is allowed
    stackReferring3 = M1(ref stackReferring2, stackReferring1);

    // this is allowed
    M2(ref stackReferring3) = stackReferring2;

    // this is NOT allowed
    M1(ref param1) = stackReferring2;

    // this is NOT allowed
    param1 = stackReferring3;

    // this is NOT allowed
    return ref stackReferring3;

    // this is allowed
    return ref param1;
}

Draft language specification

Below we describe a set of safety rules for ref-like types (ref structs) to ensure that values of these types occur only on the stack. A different, simpler set of safety rules would be possible if locals cannot be passed by reference. This specification would also permit the safe reassignment of ref locals.

Overview

We associate with each expression at compile-time the concept of what scope that expression is permitted to escape to, "safe-to-escape". Similarly, for each lvalue we maintain a concept of what scope a reference to it is permitted to escape to, "ref-safe-to-escape". For a given lvalue expression, these may be different.

These are analogous to the "safe to return" of the ref locals feature, but it is more fine-grained. Where the "safe-to-return" of an expression records only whether (or not) it may escape the enclosing method as a whole, the safe-to-escape records which scope it may escape to (which scope it may not escape beyond). The basic safety mechanism is enforced as follows. Given an assignment from an expression E1 with a safe-to-escape scope S1, to an (lvalue) expression E2 with safe-to-escape scope S2, it is an error if S2 is a wider scope than S1. By construction, the two scopes S1 and S2 are in a nesting relationship, because a legal expression is always safe-to-return from some scope enclosing the expression.

For the time being it is sufficient, for the purpose of the analysis, to support just two scopes - external to the method, and top-level scope of the method. That is because ref-like values with inner scopes cannot be created and ref locals do not support re-assignment. The rules, however, can support more than two scope levels.

The precise rules for computing the safe-to-return status of an expression, and the rules governing the legality of expressions, follow.

ref-safe-to-escape

The ref-safe-to-escape is a scope, enclosing an lvalue expression, to which it is safe for a ref to the lvalue to escape to. If that scope is the entire method, we say that a ref to the lvalue is safe to return from the method.

The ref-safe-to-escape scope for an lvalue expression can never be to a greater scope than the safe-to-escape for the same value. That means when the spec limits the safe-to-escape of a value it is implicitly also limiting the ref-safe-to-escape as well. However ref-safe-to-escape scope can be to a smaller scope than safe-to-escape. Consider that non-ref locals have safe-to-escape scope outside method but ref-safe-to-escape inside the method.

safe-to-escape

The safe-to-escape is a scope, enclosing an expression, to which it is safe for the value to escape to. If that scope is the entire method, we say that the value is safe to return from the method.

An expression whose type is not a ref struct type is always safe-to-return from the entire enclosing method. Otherwise we refer to the rules below.

Parameters

An lvalue designating a formal parameter is ref-safe-to-escape (by reference) as follows:

  • If the parameter is a ref, out, or in parameter, it is ref-safe-to-escape from the entire method (e.g. by a return ref statement); otherwise
  • If the parameter is the this parameter of a struct type, it is ref-safe-to-escape to the top-level scope of the method (but not from the entire method itself); Sample
  • Otherwise the parameter is a value parameter, and it is ref-safe-to-escape to the top-level scope of the method (but not from the method itself).

An expression that is an rvalue designating the use of a formal parameter is safe-to-escape (by value) from the entire method (e.g. by a return statement). This applies to the this parameter as well.

Locals

An lvalue designating a local variable is ref-safe-to-escape (by reference) as follows:

  • If the variable is a ref variable, then its ref-safe-to-escape is taken from the ref-safe-to-escape of its initializing expression; otherwise
  • The variable is ref-safe-to-escape the scope in which it was declared.

An expression that is an rvalue designating the use of a local variable is safe-to-escape (by value) as follows:

  • But the general rule above, a local whose type is not a ref struct type is safe-to-return from the entire enclosing method.
  • If the variable is an iteration variable of a foreach loop, then the variable's safe-to-escape scope is the same as the safe-to-escape of the foreach loop's expression.
  • A local of ref struct type and uninitialized at the point of declaration is safe-to-return from the entire enclosing method.
  • Otherwise the variable's type is a ref struct type, and the variable's declaration requires an initializer. The variable's safe-to-escape scope is the same as the safe-to-escape of its initializer.

Field reference

An lvalue designating a reference to a field, e.F, is ref-safe-to-escape (by reference) as follows:

  • If e is of a reference type, it is ref-safe-to-escape from the entire method; otherwise
  • If e is of a value type, its ref-safe-to-escape is taken from the ref-safe-to-escape of e.

An rvalue designating a reference to a field, e.F, has a safe-to-escape scope that is the same as the safe-to-escape of e.

Operators including ?:

The application of a user-defined operator is treated as a method invocation.

For an operator that yields an rvalue, such as e1 + e2 or c ? e1 : e2, the safe-to-escape of the result is the narrowest scope among the safe-to-escape of the operands of the operator. As a consequence, for a unary operator that yields an rvalue, such as +e, the safe-to-escape of the result is the safe-to-escape of the operand.

For an operator that yields an lvalue, such as c ? ref e1 : ref e2

  • the ref-safe-to-escape of the result is the narrowest scope among the ref-safe-to-escape of the operands of the operator.
  • the safe-to-escape of the operands must agree, and that is the safe-to-escape of the resulting lvalue.

Method invocation

An lvalue resulting from a ref-returning method invocation e1.M(e2, ...) is ref-safe-to-escape the smallest of the following scopes:

  • The entire enclosing method
  • the ref-safe-to-escape of all ref and out argument expressions (excluding the receiver)
  • For each in parameter of the method, if there is a corresponding expression that is an lvalue, its ref-safe-to-escape, otherwise the nearest enclosing scope
  • the safe-to-escape of all argument expressions (including the receiver)

Note: the last bullet is necessary to handle code such as

var sp = new Span(...)
return ref sp[0];

or

return ref M(sp, 0);

An rvalue resulting from a method invocation e1.M(e2, ...) is safe-to-escape from the smallest of the following scopes:

  • The entire enclosing method
  • the safe-to-escape of all argument expressions (including the receiver)

An Rvalue

An rvalue is ref-safe-to-escape from the nearest enclosing scope. This occurs for example in an invocation such as M(ref d.Length) where d is of type dynamic. It is also consistent with (and perhaps subsumes) our handling of arguments corresponding to in parameters.

Property invocations

A property invocation (either get or set) it treated as a method invocation of the underlying method by the above rules.

stackalloc

A stackalloc expression is an rvalue that is safe-to-escape to the top-level scope of the method (but not from the entire method itself).

Constructor invocations

A new expression that invokes a constructor obeys the same rules as a method invocation that is considered to return the type being constructed.

In addition safe-to-escape is no wider than the smallest of the safe-to-escape of all arguments/operands of the object initializer expressions, recursively, if initializer is present.

Span constructor

The language relies on Span<T> not having a constructor of the following form:

void Example(ref int x)
{
    // Create a span of length one
    var span = new Span<int>(ref x); 
}

Such a constructor makes Span<T> which are used as fields indistinguishable from a ref field. The safety rules described in this document depend on ref fields not being a valid construct in C# or .NET.

default expressions

A default expression is safe-to-escape from the entire enclosing method.

Language Constraints

We wish to ensure that no ref local variable, and no variable of ref struct type, refers to stack memory or variables that are no longer alive. We therefore have the following language constraints:

  • Neither a ref parameter, nor a ref local, nor a parameter or local of a ref struct type can be lifted into a lambda or local function.

  • Neither a ref parameter nor a parameter of a ref struct type may be an argument on an iterator method or an async method.

  • Neither a ref local, nor a local of a ref struct type may be in scope at the point of a yield return statement or an await expression.

  • A ref struct type may not be used as a type argument, or as an element type in a tuple type.

  • A ref struct type may not be the declared type of a field, except that it may be the declared type of an instance field of another ref struct.

  • A ref struct type may not be the element type of an array.

  • A value of a ref struct type may not be boxed:

    • There is no conversion from a ref struct type to the type object or the type System.ValueType.
    • A ref struct type may not be declared to implement any interface
    • No instance method declared in object or in System.ValueType but not overridden in a ref struct type may be called with a receiver of that ref struct type.
    • No instance method of a ref struct type may be captured by method conversion to a delegate type.

  • For a ref reassignment e1 = ref e2, the ref-safe-to-escape of e2 must be at least as wide a scope as the ref-safe-to-escape of e1.

  • For a ref return statement return ref e1, the ref-safe-to-escape of e1 must be ref-safe-to-escape from the entire method. (TODO: Do we also need a rule that e1 must be safe-to-escape from the entire method, or is that redundant?)

  • For a return statement return e1, the safe-to-escape of e1 must be safe-to-escape from the entire method.

  • For an assignment e1 = e2, if the type of e1 is a ref struct type, then the safe-to-escape of e2 must be at least as wide a scope as the safe-to-escape of e1.

  • For a method invocation if there is a ref or out argument of a ref struct type (including the receiver unless the type is readonly), with safe-to-escape E1, then no argument (including the receiver) may have a narrower safe-to-escape than E1. Sample

  • A local function or anonymous function may not refer to a local or parameter of ref struct type declared in an enclosing scope.

Open Issue: We need some rule that permits us to produce an error when needing to spill a stack value of a ref struct type at an await expression, for example in the code

Foo(new Span<int>(...), await e2);

Explanations

These explanations and samples help explain why many of the safety rules above exist

Method Arguments Must Match

When invoking a method where there is an out or ref parameter that is a ref struct then all of the ref struct parameters need to have the same lifetime. This is necessary because C# must make all of its decisions around lifetime safety based on the information available in the signature of the method and the lifetime of the values at the call site.

When there are ref parameters that are ref struct then there is the potential that they could swap around their contents. Hence at the call site we must ensure all of these potential swaps are compatible. If the language didn't enforce that then it will allow for bad code like the following.

void M1(ref Span<int> s1)
{
    Span<int> s2 = stackalloc int[1];
    Swap(ref s1, ref s2);
}

void Swap(ref Span<int> x, ref Span<int> y)
{
    // This will effectively assign the stackalloc to the s1 parameter and allow it
    // to escape to the caller of M1
    ref x = ref y; 
}

This analysis of ref parameters includes the receiver in instance methods. This is necessary because it can be used to store values passed in as parameters, just as a ref parameter could`. This means with mismatched lifetimes you could create a type safety hole in the following way:

ref struct S
{
    public Span<int> Span;

    public void Set(Span<int> span)
    {
        Span = span;
    }
}

void Broken(ref S s)
{
    Span<int> span = stackalloc int[1];

    // The result of a stackalloc is now stored in s.Span and escaped to the caller
    // of Broken
    s.Set(span); 
}

For the purpose of this analysis the receiver is considered an in, not a ref, if the type is a readonly struct. In that case the receiver cannot be used to store values from other parameters, it is effectively an in parameter for analysis purposes. Hence the same example above is legal when S is readonly because the span cannot be stored anywhere.

Struct This Escape

When it comes to span safety rules, the this value in an instance member is modeled as a parameter to the member. Now for a struct the type of this is actually ref S where in a class it's simply S (for members of a class / struct named S).

Yet this has different escaping rules than other ref parameters. Specifically it is not ref-safe-to-escape while other parameters are:

ref struct S
{ 
    int Field;

    // Illegal because `this` isn't safe to escape as ref
    ref int Get() => ref Field;

    // Legal
    ref int GetParam(ref int p) => ref p;
}

The reason for this restriction actually has little to do with struct member invocation. There are some rules that need to be worked out with respect to member invocation on struct members where the receiver is an rvalue. But that is very approachable.

The reason for this restriction is actually about interface invocation. Specifically it comes down to whether or not the following sample should or should not compile;

interface I1
{
    ref int Get();
}

ref int Use<T>(T p)
    where T : I1
{
    return ref p.Get();
}

Consider the case where T is instantiated as a struct. If the this parameter is ref-safe-to-escape then the return of p.Get could point to the stack (specifically it could be a field inside of the instantiated type of T). That means the language could not allow this sample to compile as it could be returning a ref to a stack location. On the other hand if this is not ref-safe-to-escape then p.Get cannot refer to the stack and hence it's safe to return.

This is why the escapability of this in a struct is really all about interfaces. It can absolutely be made to work but it has a trade off. The design eventually came down in favor of making interfaces more flexible.

There is potential for us to relax this in the future though.

Future Considerations

Length one Span<T> over ref values

Though not legal today there are cases where creating a length one Span<T> instance over a value would be beneficial:

void RefExample()
{
    int x = ...;

    // Today creating a length one Span<int> requires a stackalloc and a new 
    // local
    Span<int> span1 = stackalloc [] { x };
    Use(span1);
    x = span1[0]; 

    // Simpler to just allow length one span
    var span2 = new Span<int>(ref x);
    Use(span2);
}

This feature gets more compelling if we lift the restrictions on fixed sized buffers as it would allow for Span<T> instances of even greater length.

If there is ever a need to go down this path then the language could accommodate this by ensuring such Span<T> instances were downward facing only. That is they were only ever safe-to-escape to the scope in which they were created. This ensure the language never had to consider a ref value escaping a method via a ref struct return or field of ref struct. This would likely also require further changes to recognize such constructors as capturing a ref parameter in this way though.