Re: [PATCH v12 2/3] rust: io: mem: add a generic iomem abstraction

["Danilo Krummrich" <dakr@kernel.org>]

On Mon Jul 14, 2025 at 1:51 PM CEST, Alice Ryhl wrote:
> On Thu, Jul 10, 2025 at 10:58:13AM -0300, Daniel Almeida wrote:
>> Hi Alice,
>> 
>> >> +impl<const SIZE: usize> IoMem<SIZE> {
>> >> +    fn ioremap(resource: &Resource) -> Result<Self> {
>> >> +        let size = resource.size();
>> >> +        if size == 0 {
>> >> +            return Err(EINVAL);
>> >> +        }
>> >> +
>> >> +        let res_start = resource.start();
>> >> +
>> >> +        let addr = if resource
>> >> +            .flags()
>> >> +            .contains(io::resource::flags::IORESOURCE_MEM_NONPOSTED)
>> >> +        {
>> >> +            // SAFETY:
>> >> +            // - `res_start` and `size` are read from a presumably valid `struct resource`.
>> >> +            // - `size` is known not to be zero at this point.
>> >> +            unsafe { bindings::ioremap_np(res_start, size as usize) }
>> > 
>> > Here you cast from ResourceSize to usize. Are you sure that is correct?
>> > I thought those types could be different.
>> 
>> This seems to what C is doing as well, i.e.:
>> 
>> static void __iomem *__devm_ioremap(struct device *dev, resource_size_t offset,
>> 				    resource_size_t size, <---------
>> 				    enum devm_ioremap_type type)
>> {
>> 
>>         […]
>> 
>> 	case DEVM_IOREMAP_NP:
>> 		addr = ioremap_np(offset, size);
>> 		break;
>> 	}
>> 
>> 
>> Where:
>> 
>> `static inline void __iomem *ioremap_np(phys_addr_t offset, size_t size)`
>> 
>> IOW: this stems from the mix and match of types used the C API itself.
>> 
>> What do you suggest here? Maybe a try_into() then?
>
> What a mess.

Yeah, it mixes up types describing CPU word width and bus address width. :(

> It looks like there aren't any 32-bit architectures that
> define ioremap_np. This means that sometimes this cast will be lossy,
> but in those cases the function body just returns NULL and doesn't read
> the size.
>
> I would probably cast to an underscore instead of explicitly mentioning
> the target type and make a comment about it.

I think fixing up the C side would be even nicer, but for the scope of this
series that's fine. The comment should mention that, ultimately, we want to fix
up the C side type wise.