[ORLinux] Linux device access

[Stefan Kristiansson]

On 4.7.2011 7:22, Jonas Bonn wrote:
> On Sun, 2011-07-03 at 23:39 +0300, Stefan Kristiansson wrote:
>> Is there any particular reason for not using the wishbone wrappers instead of having the ifdefs in
>> the driver code?
>> I guess it's a matter of taste and I'm fine with having them in the drivers code, but I kind of liked the
>> wishbone wrappers approach better.
>>
>
> The problem, as I see it, is that the hardware isn't necessarily
> wishbone-only.  I could be  wrong (in particular for the ethmac case),
> but I believe there are #ifdefs in the RTL that allow you to put the
> device on an Avalon bus.  Using wishbone-specific accessors in the
> driver wouldn't really be right for people using the driver on other
> buses... 
>

Those ifdefs are just for some internal workaround for people creating an avalon<->wishbone wrapper.
The external interface of the core is still wishbone.

> If we do a wrapper, we still end up with:
>
> #ifdef CONFIG_WISHBONE_BUS
> 	wb_ioread32(...)
> #else
> 	ioread32(...)
> #endif
>
> ...so if we need to do an #ifdef anyway in order to specify the bus,
> then we might as well just do it for endianess.
>
> There is one other solution and that's to embed information about the
> device endianess into the __iomem cookie returned by ioremap.
> Unfortunately, doing that changes the semantics of ioread32 and also
> makes it less of a lightweight instruction than it is today.
>
> There's still an argument for putting some kind of wrapper back into
> io.h, but at this point in time, I'm not quite sure how to do it.
> Suggestions welcome.
>

I don't have any concrete suggestions at the moment,
so I'd say having the ifdefs in the drivers is good enough for the time being.

Stefan