Re: [PATCH 09/22] rust: pin-init: move impl `Zeroable` for `Opaque` and `Option<KBox<T>>` into the kernel crate

[Benno Lossin]

On Wed Mar 5, 2025 at 12:26 PM CET, Andreas Hindborg wrote:
> "Benno Lossin" <benno.lossin@xxxxxxxxx> writes:
>
>> In order to make pin-init a standalone crate, move kernel-specific code
>> directly into the kernel crate. Since `Opaque<T>` and `KBox<T>` are part
>> of the kernel, move their `Zeroable` implementation into the kernel
>> crate.
>>
>> Signed-off-by: Benno Lossin <benno.lossin@xxxxxxxxx>
>> ---
>>  rust/kernel/alloc/kbox.rs | 8 +++++++-
>>  rust/kernel/types.rs      | 5 ++++-
>>  rust/pin-init/src/lib.rs  | 8 +-------
>>  3 files changed, 12 insertions(+), 9 deletions(-)
>>
>> diff --git a/rust/kernel/alloc/kbox.rs b/rust/kernel/alloc/kbox.rs
>> index 39a3ea7542da..9861433559dc 100644
>> --- a/rust/kernel/alloc/kbox.rs
>> +++ b/rust/kernel/alloc/kbox.rs
>> @@ -15,7 +15,7 @@
>>  use core::ptr::NonNull;
>>  use core::result::Result;
>>
>> -use crate::init::{InPlaceWrite, Init, PinInit};
>> +use crate::init::{InPlaceWrite, Init, PinInit, Zeroable};
>>  use crate::init_ext::InPlaceInit;
>>  use crate::types::ForeignOwnable;
>>
>> @@ -100,6 +100,12 @@
>>  /// ```
>>  pub type KVBox<T> = Box<T, super::allocator::KVmalloc>;
>>
>> +// SAFETY: All zeros is equivalent to `None` (option layout optimization guarantee).
>> +//
>> +// In this case we are allowed to use `T: ?Sized`, since all zeros is the `None` variant and there
>> +// is no problem with a VTABLE pointer being null.
>> +unsafe impl<T: ?Sized, A: Allocator> Zeroable for Option<Box<T, A>> {}
>
> Could you elaborate the statement related to vtable pointers? How does
> that come into play for `Option<Box<_>>`? Is it for fat pointers to
> trait objects?

Yes it is for fat pointers, if you have a `x: *mut dyn Trait`, then you
aren't allowed to write all zeroes to `x`, because the VTABLE pointer
(that is part of the fat pointer) is not allowed to be null.

Now for `Option<Box<_>>`, this doesn't matter, as there if the normal
pointer part of the fat pointer is all zeroes, then the VTABLE pointer
part is considered padding bytes, as it's the `None` variant.

---
Cheers,
Benno