แก้ไข

แชร์ผ่าน

Native interoperability best practices

  • บทความ

.NET gives you various ways to customize your native interoperability code. This article includes the guidance that Microsoft's .NET teams follow for native interoperability.

General guidance

The guidance in this section applies to all interop scenarios.

  • ✔️ DO use [LibraryImport], if possible, when targeting .NET 7+.
    • There are cases when using [DllImport] is appropriate. A code analyzer with ID SYSLIB1054 tells you when that's the case.
  • ✔️ DO use the same naming and capitalization for your methods and parameters as the native method you want to call.
  • ✔️ CONSIDER using the same naming and capitalization for constant values.
  • ✔️ DO use .NET types that map closest to the native type. For example, in C#, use uint when the native type is unsigned int.
  • ✔️ DO prefer expressing higher level native types using .NET structs rather than classes.
  • ✔️ DO prefer using function pointers, as opposed to Delegate types, when passing callbacks to unmanaged functions in C#.
  • ✔️ DO use [In] and [Out] attributes on array parameters.
  • ✔️ DO only use [In] and [Out] attributes on other types when the behavior you want differs from the default behavior.
  • ✔️ CONSIDER using System.Buffers.ArrayPool<T> to pool your native array buffers.
  • ✔️ CONSIDER wrapping your P/Invoke declarations in a class with the same name and capitalization as your native library.
    • This allows your [LibraryImport] or [DllImport] attributes to use the C# nameof language feature to pass in the name of the native library and ensure that you didn't misspell the name of the native library.
  • ✔️ DO use SafeHandle handles to manage lifetime of objects that encapsulate unmanaged resources. For more information, see Cleaning up unmanaged resources.
  • ❌ AVOID finalizers to manage lifetime of objects that encapsulate unmanaged resources. For more information, see Implement a Dispose method.

LibraryImport attribute settings

A code analyzer, with ID SYSLIB1054, helps guide you with LibraryImportAttribute. In most cases, the use of LibraryImportAttribute requires an explicit declaration rather than relying on default settings. This design is intentional and helps avoid unintended behavior in interop scenarios.

DllImport attribute settings

Setting Default Recommendation Details
PreserveSig true Keep default When this is explicitly set to false, failed HRESULT return values will be turned into exceptions (and the return value in the definition becomes null as a result).
SetLastError false Depends on the API Set this to true if the API uses GetLastError and use Marshal.GetLastWin32Error to get the value. If the API sets a condition that says it has an error, get the error before making other calls to avoid inadvertently having it overwritten.
CharSet Compiler-defined (specified in the charset documentation) Explicitly use CharSet.Unicode or CharSet.Ansi when strings or characters are present in the definition This specifies marshalling behavior of strings and what ExactSpelling does when false. Note that CharSet.Ansi is actually UTF8 on Unix. Most of the time Windows uses Unicode while Unix uses UTF8. See more information on the documentation on charsets.
ExactSpelling false true Set this to true and gain a slight perf benefit as the runtime will not look for alternate function names with either an "A" or "W" suffix depending on the value of the CharSet setting ("A" for CharSet.Ansi and "W" for CharSet.Unicode).

String parameters

A string is pinned and used directly by native code (rather than copied) when passed by value (not ref or out) and any one of the following:

❌ DON'T use [Out] string parameters. String parameters passed by value with the [Out] attribute can destabilize the runtime if the string is an interned string. See more information about string interning in the documentation for String.Intern.

✔️ CONSIDER char[] or byte[] arrays from an ArrayPool when native code is expected to fill a character buffer. This requires passing the argument as [Out].

DllImport-specific guidance

✔️ CONSIDER setting the CharSet property in [DllImport] so the runtime knows the expected string encoding.

✔️ CONSIDER avoiding StringBuilder parameters. StringBuilder marshalling always creates a native buffer copy. As such, it can be extremely inefficient. Take the typical scenario of calling a Windows API that takes a string:

  1. Create a StringBuilder of the desired capacity (allocates managed capacity) {1}.
  2. Invoke:
    1. Allocates a native buffer {2}.
    2. Copies the contents if [In] (the default for a StringBuilder parameter).
    3. Copies the native buffer into a newly allocated managed array if [Out] {3} (also the default for StringBuilder).
  3. ToString() allocates yet another managed array {4}.

That's {4} allocations to get a string out of native code. The best you can do to limit this is to reuse the StringBuilder in another call, but this still only saves one allocation. It's much better to use and cache a character buffer from ArrayPool. You can then get down to just the allocation for the ToString() on subsequent calls.

The other issue with StringBuilder is that it always copies the return buffer back up to the first null. If the passed back string isn't terminated or is a double-null-terminated string, your P/Invoke is incorrect at best.

If you do use StringBuilder, one last gotcha is that the capacity does not include a hidden null, which is always accounted for in interop. It's common for people to get this wrong as most APIs want the size of the buffer including the null. This can result in wasted/unnecessary allocations. Additionally, this gotcha prevents the runtime from optimizing StringBuilder marshalling to minimize copies.

For more information on string marshalling, see Default Marshalling for Strings and Customizing string marshalling.

Windows Specific For [Out] strings the CLR will use CoTaskMemFree by default to free strings or SysStringFree for strings that are marked as UnmanagedType.BSTR. For most APIs with an output string buffer: The passed in character count must include the null. If the returned value is less than the passed in character count the call has succeeded and the value is the number of characters without the trailing null. Otherwise the count is the required size of the buffer including the null character.

  • Pass in 5, get 4: The string is 4 characters long with a trailing null.
  • Pass in 5, get 6: The string is 5 characters long, need a 6 character buffer to hold the null. Windows Data Types for Strings

Boolean parameters and fields

Booleans are easy to mess up. By default, a .NET bool is marshalled to a Windows BOOL, where it's a 4-byte value. However, the _Bool, and bool types in C and C++ are a single byte. This can lead to hard to track down bugs as half the return value will be discarded, which will only potentially change the result. For more for information on marshalling .NET bool values to C or C++ bool types, see the documentation on customizing boolean field marshalling.

GUIDs

GUIDs are usable directly in signatures. Many Windows APIs take GUID& type aliases like REFIID. When the method signature contains a reference parameter, place either a ref keyword or a [MarshalAs(UnmanagedType.LPStruct)] attribute on the GUID parameter declaration.

GUID By-ref GUID
KNOWNFOLDERID REFKNOWNFOLDERID

❌ DON'T Use [MarshalAs(UnmanagedType.LPStruct)] for anything other than ref GUID parameters.

Blittable types

Blittable types are types that have the same bit-level representation in managed and native code. As such they do not need to be converted to another format to be marshalled to and from native code, and as this improves performance they should be preferred. Some types are not blittable but are known to contain blittable contents. These types have similar optimizations as blittable types when they are not contained in another type, but are not considered blittable when in fields of structs or for the purposes of UnmanagedCallersOnlyAttribute.

Blittable types when runtime marshalling is enabled

Blittable types:

  • byte, sbyte, short, ushort, int, uint, long, ulong, single, double
  • structs with fixed layout that only have blittable value types for instance fields
    • fixed layout requires [StructLayout(LayoutKind.Sequential)] or [StructLayout(LayoutKind.Explicit)]
    • structs are LayoutKind.Sequential by default

Types with blittable contents:

  • non-nested, one-dimensional arrays of blittable primitive types (for example, int[])
  • classes with fixed layout that only have blittable value types for instance fields
    • fixed layout requires [StructLayout(LayoutKind.Sequential)] or [StructLayout(LayoutKind.Explicit)]
    • classes are LayoutKind.Auto by default

NOT blittable:

  • bool

SOMETIMES blittable:

  • char

Types with SOMETIMES blittable contents:

  • string

When blittable types are passed by reference with in, ref, or out, or when types with blittable contents are passed by value, they're simply pinned by the marshaller instead of being copied to an intermediate buffer.

char is blittable in a one-dimensional array or if it's part of a type that contains it's explicitly marked with [StructLayout] with CharSet = CharSet.Unicode.

[StructLayout(LayoutKind.Sequential, CharSet = CharSet.Unicode)]
public struct UnicodeCharStruct
{
    public char c;
}

string contains blittable contents if it isn't contained in another type and is being passed by value (not ref or out) as an argument and any one of the following:

You can see if a type is blittable or contains blittable contents by attempting to create a pinned GCHandle. If the type isn't a string or considered blittable, GCHandle.Alloc will throw an ArgumentException.

Blittable types when runtime marshalling is disabled

When runtime marshalling is disabled, the rules for which types are blittable are significantly simpler. All types that are C# unmanaged types and don't have any fields that are marked with [StructLayout(LayoutKind.Auto)] are blittable. All types that are not C# unmanaged types are not blittable. The concept of types with blittable contents, such as arrays or strings, does not apply when runtime marshalling is disabled. Any type that is not considered blittable by the aforementioned rule is unsupported when runtime marshalling is disabled.

These rules differ from the built-in system primarily in situations where bool and char are used. When marshalling is disabled, bool is passed as a 1-byte value and not normalized and char is always passed as a 2-byte value. When runtime marshalling is enabled, bool can map to a 1, 2, or 4-byte value and is always normalized, and char maps to either a 1 or 2-byte value depending on the CharSet.

✔️ DO make your structures blittable when possible.

For more information, see:

Keeping managed objects alive

GC.KeepAlive() will ensure an object stays in scope until the KeepAlive method is hit.

HandleRef allows the marshaller to keep an object alive for the duration of a P/Invoke. It can be used instead of IntPtr in method signatures. SafeHandle effectively replaces this class and should be used instead.

GCHandle allows pinning a managed object and getting the native pointer to it. The basic pattern is:

GCHandle handle = GCHandle.Alloc(obj, GCHandleType.Pinned);
IntPtr ptr = handle.AddrOfPinnedObject();
handle.Free();

Pinning isn't the default for GCHandle. The other major pattern is for passing a reference to a managed object through native code and back to managed code, usually with a callback. Here is the pattern:

GCHandle handle = GCHandle.Alloc(obj);
SomeNativeEnumerator(callbackDelegate, GCHandle.ToIntPtr(handle));

// In the callback
GCHandle handle = GCHandle.FromIntPtr(param);
object managedObject = handle.Target;

// After the last callback
handle.Free();

Don't forget that GCHandle needs to be explicitly freed to avoid memory leaks.

Common Windows data types

Here is a list of data types commonly used in Windows APIs and which C# types to use when calling into the Windows code.

The following types are the same size on 32-bit and 64-bit Windows, despite their names.

Width Windows C# Alternative
32 BOOL int bool
8 BOOLEAN byte [MarshalAs(UnmanagedType.U1)] bool
8 BYTE byte
8 UCHAR byte
8 UINT8 byte
8 CCHAR byte
8 CHAR sbyte
8 CHAR sbyte
8 INT8 sbyte
16 CSHORT short
16 INT16 short
16 SHORT short
16 ATOM ushort
16 UINT16 ushort
16 USHORT ushort
16 WORD ushort
32 INT int
32 INT32 int
32 LONG int See CLong and CULong.
32 LONG32 int
32 CLONG uint See CLong and CULong.
32 DWORD uint See CLong and CULong.
32 DWORD32 uint
32 UINT uint
32 UINT32 uint
32 ULONG uint See CLong and CULong.
32 ULONG32 uint
64 INT64 long
64 LARGE_INTEGER long
64 LONG64 long
64 LONGLONG long
64 QWORD long
64 DWORD64 ulong
64 UINT64 ulong
64 ULONG64 ulong
64 ULONGLONG ulong
64 ULARGE_INTEGER ulong
32 HRESULT int
32 NTSTATUS int

The following types, being pointers, do follow the width of the platform. Use IntPtr/UIntPtr for these.

Signed Pointer Types (use IntPtr) Unsigned Pointer Types (use UIntPtr)
HANDLE WPARAM
HWND UINT_PTR
HINSTANCE ULONG_PTR
LPARAM SIZE_T
LRESULT
LONG_PTR
INT_PTR

A Windows PVOID, which is a C void*, can be marshalled as either IntPtr or UIntPtr, but prefer void* when possible.

Windows Data Types

Data Type Ranges

Formerly built-in supported types

There are rare instances when built-in support for a type is removed.

The UnmanagedType.HString and UnmanagedType.IInspectable built-in marshal support was removed in the .NET 5 release. You must recompile binaries that use this marshalling type and that target a previous framework. It's still possible to marshal this type, but you must marshal it manually, as the following code example shows. This code will work moving forward and is also compatible with previous frameworks.

public sealed class HStringMarshaler : ICustomMarshaler
{
    public static readonly HStringMarshaler Instance = new HStringMarshaler();

    public static ICustomMarshaler GetInstance(string _) => Instance;

    public void CleanUpManagedData(object ManagedObj) { }

    public void CleanUpNativeData(IntPtr pNativeData)
    {
        if (pNativeData != IntPtr.Zero)
        {
            Marshal.ThrowExceptionForHR(WindowsDeleteString(pNativeData));
        }
    }

    public int GetNativeDataSize() => -1;

    public IntPtr MarshalManagedToNative(object ManagedObj)
    {
        if (ManagedObj is null)
            return IntPtr.Zero;

        var str = (string)ManagedObj;
        Marshal.ThrowExceptionForHR(WindowsCreateString(str, str.Length, out var ptr));
        return ptr;
    }

    public object MarshalNativeToManaged(IntPtr pNativeData)
    {
        if (pNativeData == IntPtr.Zero)
            return null;

        var ptr = WindowsGetStringRawBuffer(pNativeData, out var length);
        if (ptr == IntPtr.Zero)
            return null;

        if (length == 0)
            return string.Empty;

        return Marshal.PtrToStringUni(ptr, length);
    }

    [DllImport("api-ms-win-core-winrt-string-l1-1-0.dll")]
    [DefaultDllImportSearchPaths(DllImportSearchPath.System32)]
    private static extern int WindowsCreateString([MarshalAs(UnmanagedType.LPWStr)] string sourceString, int length, out IntPtr hstring);

    [DllImport("api-ms-win-core-winrt-string-l1-1-0.dll")]
    [DefaultDllImportSearchPaths(DllImportSearchPath.System32)]
    private static extern int WindowsDeleteString(IntPtr hstring);

    [DllImport("api-ms-win-core-winrt-string-l1-1-0.dll")]
    [DefaultDllImportSearchPaths(DllImportSearchPath.System32)]
    private static extern IntPtr WindowsGetStringRawBuffer(IntPtr hstring, out int length);
}

// Example usage:
[DllImport("api-ms-win-core-winrt-l1-1-0.dll", PreserveSig = true)]
internal static extern int RoGetActivationFactory(
    /*[MarshalAs(UnmanagedType.HString)]*/[MarshalAs(UnmanagedType.CustomMarshaler, MarshalTypeRef = typeof(HStringMarshaler))] string activatableClassId,
    [In] ref Guid iid,
    [Out, MarshalAs(UnmanagedType.IUnknown)] out object factory);

Cross-platform data type considerations

There are types in the C/C++ language that have latitude in how they are defined. When writing cross-platform interop, cases can arise where platforms differ and can cause issues if not considered.

C/C++ long

C/C++ long and C# long are not necessarily the same size.

The long type in C/C++ is defined to have "at least 32" bits. This means there is a minimum number of required bits, but platforms can choose to use more bits if desired. The following table illustrates the differences in provided bits for the C/C++ long data type between platforms.

Platform 32-bit 64-bit
Windows 32 32
macOS/*nix 32 64

In contrast, C# long is always 64 bit. For this reason, it's best to avoid using C# long to interop with C/C++ long.

(This problem with C/C++ long does not exist for C/C++ char, short, int, and long long as they are 8, 16, 32, and 64 bits respectively on all of these platforms.)

In .NET 6 and later versions, use the CLong and CULong types for interop with C/C++ long and unsigned long data types. The following example is for CLong, but you can use CULong to abstract unsigned long in a similar way.

// Cross platform C function
// long Function(long a);
[DllImport("NativeLib")]
extern static CLong Function(CLong a);

// Usage
nint result = Function(new CLong(10)).Value;

When targeting .NET 5 and earlier versions, you should declare separate Windows and non-Windows signatures to handle the problem.

static readonly bool IsWindows = RuntimeInformation.IsOSPlatform(OSPlatform.Windows);

// Cross platform C function
// long Function(long a);

[DllImport("NativeLib", EntryPoint = "Function")]
extern static int FunctionWindows(int a);

[DllImport("NativeLib", EntryPoint = "Function")]
extern static nint FunctionUnix(nint a);

// Usage
nint result;
if (IsWindows)
{
    result = FunctionWindows(10);
}
else
{
    result = FunctionUnix(10);
}

Structs

Managed structs are created on the stack and aren't removed until the method returns. By definition then, they are "pinned" (it won't get moved by the GC). You can also simply take the address in unsafe code blocks if native code won't use the pointer past the end of the current method.

Blittable structs are much more performant as they can simply be used directly by the marshalling layer. Try to make structs blittable (for example, avoid bool). For more information, see the Blittable Types section.

If the struct is blittable, use sizeof() instead of Marshal.SizeOf<MyStruct>() for better performance. As mentioned above, you can validate that the type is blittable by attempting to create a pinned GCHandle. If the type is not a string or considered blittable, GCHandle.Alloc will throw an ArgumentException.

Pointers to structs in definitions must either be passed by ref or use unsafe and *.

✔️ DO match the managed struct as closely as possible to the shape and names that are used in the official platform documentation or header.

✔️ DO use the C# sizeof() instead of Marshal.SizeOf<MyStruct>() for blittable structures to improve performance.

❌ AVOID using classes to express complex native types through inheritance.

❌ AVOID using System.Delegate or System.MulticastDelegate fields to represent function pointer fields in structures.

Since System.Delegate and System.MulticastDelegate don't have a required signature, they don't guarantee that the delegate passed in will match the signature the native code expects. Additionally, in .NET Framework and .NET Core, marshalling a struct containing a System.Delegate or System.MulticastDelegate from its native representation to a managed object can destabilize the runtime if the value of the field in the native representation isn't a function pointer that wraps a managed delegate. In .NET 5 and later versions, marshalling a System.Delegate or System.MulticastDelegate field from a native representation to a managed object is not supported. Use a specific delegate type instead of System.Delegate or System.MulticastDelegate.

Fixed Buffers

An array like INT_PTR Reserved1[2] has to be marshalled to two IntPtr fields, Reserved1a and Reserved1b. When the native array is a primitive type, we can use the fixed keyword to write it a little more cleanly. For example, SYSTEM_PROCESS_INFORMATION looks like this in the native header:

typedef struct _SYSTEM_PROCESS_INFORMATION {
    ULONG NextEntryOffset;
    ULONG NumberOfThreads;
    BYTE Reserved1[48];
    UNICODE_STRING ImageName;
...
} SYSTEM_PROCESS_INFORMATION

In C#, we can write it like this:

internal unsafe struct SYSTEM_PROCESS_INFORMATION
{
    internal uint NextEntryOffset;
    internal uint NumberOfThreads;
    private fixed byte Reserved1[48];
    internal Interop.UNICODE_STRING ImageName;
    ...
}

However, there are some gotchas with fixed buffers. Fixed buffers of non-blittable types won't be correctly marshalled, so the in-place array needs to be expanded out to multiple individual fields. Additionally, in .NET Framework and .NET Core before 3.0, if a struct containing a fixed buffer field is nested within a non-blittable struct, the fixed buffer field won't be correctly marshalled to native code.