Using C structures with Flexible Array Members

Since time immemorial, C programmers have been using what was called "the struct hack". This is a technique for packing a fixed-size structure and a variable-sized tail within the same memory allocation. Typically this looks like:

struct MyRecord {
    time_t timestamp;
    unsigned seq;
    size_t len;
    char payload[0];
};

Because this is so useful, it was standardized in C99 as "flexible array members", using almost identical syntax:

struct MyRecord {
    time_t timestamp;
    unsigned seq;
    size_t len;
    char payload[]; // NOTE: empty []
};

Bindgen supports these structures in two different ways.

__IncompleteArrayField

By default, bindgen will generate the corresponding Rust structure:

#[repr(C)]
struct MyRecord {
    pub timestamp: time_t,
    pub seq: ::std::os::raw::c_uint,
    pub len: usize,
    pub payload: __IncompleteArrayField<::std::os::raw::c_char>,
}

The __IncompleteArrayField type is zero-sized, so this structure represents the prefix without any trailing data. In order to access that data, it provides the as_slice unsafe method:

    // SAFETY: there's at least `len` bytes allocated and initialized after `myrecord`
    let payload = unsafe { myrecord.payload.as_slice(myrecord.len) };

There's also as_mut_slice which does the obvious.

These are unsafe simply because it's up to you to provide the right length (in elements of whatever type payload is) as there's no way for Rust or Bindgen to know. In this example, the length is a very straightforward len field in the structure, but it could be encoded in any number of ways within the structure, or come from somewhere else entirely.

One big caveat with this technique is that std::mem::size_of (or size_of_val) will only include the size of the prefix structure. if you're working out how much storage the whole structure is using, you'll need to add the suffix yourself.

Using Dynamically Sized Types

If you invoke bindgen with the --flexarray-dst option, it will generate something not quite like this:

#[repr(C)]
struct MyRecord {
    pub timestamp: time_t,
    pub seq: ::std::os::raw::c_uint,
    pub len: usize,
    pub payload: [::std::os::raw::c_char],
}

Rust has a set of types which are almost exact analogs for these Flexible Array Member types: the Dynamically Sized Type ("DST").

This looks almost identical to a normal Rust structure, except that you'll note the type of the payload field is a raw slice [...] rather than the usual reference to slice &[...].

That payload: [c_char] is telling Rust that it can't directly know the total size of this structure - the payload field takes an amount of space that's determined at runtime. This means you can't directly use values of this type, only references: &MyRecord.

In practice, this is very awkward. So instead, bindgen generates:

#[repr(C)]
struct MyRecord<FAM: ?Sized = [::std::os::raw::c_char; 0]> {
    pub timestamp: time_t,
    pub seq: ::std::os::raw::c_uint,
    pub len: usize,
    pub payload: FAM,
}

That is:

  1. a type parameter FAM which represents the type of the payload field,
  2. it's ?Sized meaning it can be unsized (ie, a DST)
  3. it has the default type of [c_char; 0] - that is a zero-sized array of characters

This means that referencing plain MyRecord will be exactly like MyRecord with __IncompleteArrayField: it is a fixed-sized structure which you can manipulate like a normal Rust value.

But how do you get to the DST part?

Bindgen will also implement a set of helper methods for this:

// Static sized variant
impl MyRecord<[::std::os::raw::c_char; 0]> {
    pub unsafe fn flex_ref(&self, len: usize) -> &MyRecord<[::std::os::raw::c_char]> { ... }
    pub unsafe fn flex_mut_ref(&mut self, len: usize) -> &mut MyRecord<[::std::os::raw::c_char]> { ... }
    // And some raw pointer variants
}

These will take a sized MyRecord<[c_char; 0]> and a length in elements, and return a reference to a DST MyRecord<[c_char]> where the payload field is a fully usable slice of len characters.

The magic here is that the reference is a fat pointer, which not only encodes the address, but also the dynamic size of the final field, just like a reference to a slice is. This means that you get full bounds checked access to the payload field like any other Rust slice.

It also means that doing mem::size_of_val(myrecord) will return the complete size of this structure, including the suffix.

You can go the other way:

// Dynamic sized variant
impl MyRecord<[::std::os::raw::c_char]> {
    pub fn fixed(&self) -> (&MyRecord<[::std::os::raw::c_char; 0]>, usize) { ... }
    pub fn fixed_mut(&mut self) -> (&mut MyRecord<[::std::os::raw::c_char; 0]>, usize) { ... }
    pub fn layout(len: usize) -> std::alloc::Layout { ... }
}

which takes the DST variant of the structure and returns the sized variant, along with the number of elements are after it. These are all completely safe because all the information needed is part of the fat &self reference.

The layout function takes a length and returns the Layout - that is, size and alignment, so that you can allocate memory for the structure (for example, using malloc so you can pass it to a C function).

Unfortunately the language features needed to support these methods are still unstable:

As a result, if you don't specify --rust-target nightly you'll just get the bare type definitions, but no real way to use them. It's often convenient to add the

--raw-line '#![feature(ptr_metadata,layout_for_ptr)]'

option if you're generating Rust as a stand-alone crate. Otherwise you'll need to add the feature line to your containing crate.