[Qemu-devel] Re: [RFC] API change for pci_set_word and related functions

[Anthony Liguori]

On 01/11/2010 04:10 PM, Stefan Weil wrote:
> > I haven't read this whole thread, but I really prefer things like
> >
> > pci_set_vendor_id(pci_dev, XXXX);
> >
> > A close alternative, would be some refactoring to allow PCI config
> > space to be represented as a C structure.  Gerd had some patches at
> > one point for this.
> >
> > Regards,
> >
> > Anthony Liguori
> >      
> This is a good solution for the standard configuration entries,
> so most code could use such calls if they were complete.
>
> For entries above offset 0x40, I'm afraid that it won't work
> (neither with individual functions nor with a C structure).
>    

Which is fine. Device specific entries in the config space are uncommon 
compared to the standard entries.  The same model could apply (devices 
write their own wrapper functions).

The problem with the set_word interface is that quite a lot of important 
config space fields are stored in a byte or less.

Regards,

Anthony Liguori

> Regards,
>
> Stefan Weil
>
>