On Sat, Sep 22, 2007 at 12:25:51AM -0600, Matthew Wilcox wrote:
> On Fri, Sep 21, 2007 at 11:39:36PM +0100, Alan Cox wrote:
> > On Fri, 21 Sep 2007 17:15:16 -0500
> > Olof Johansson <[EMAIL PROTECTED]> wrote:
> >
> > > Convert the io_req_t members to kio_addr_t, to allow use on machines with
> > > more than 16 bits worth of IO ports (i.e. secondary busses on ppc64, etc).
> >
> > What about the formatting and field widths ?
> >
> > ulong would probably be a lot saner than kio_addr_t and yet more type
> > obfuscation.
>
> I don't think anyone uses ioports > 32bit. Certainly i386 takes an int
> port as parameter to {in,out}[bwl] (and it really only uses 16-bits).
> parisc uses 24 bits. I don't know what the various ppcs do, but pci
> bars can only be 32-bit for ioports. So my opinion is that ioports
> should be uint, not ulong.
PPC would do just fine with 32-bit as well, which is what I wanted in
the first place. I just went with the local coding standard of pcmcia
and switched to kio_addr_t.
I suppose it's a janitorial todo item but with the maintainer MIA I
don't want to mess around with it too much, since I can't really test
much besides the PPC stuff I have.
As for the formatting/padding widths: Some platforms had ioaddr_t's
that were 32 bit already, so it was already broken on those, and the
only drawback is missing 0-padding. It'd look a bit silly to pad to 16
0:s anyway at the moment, so I think I'd prefer to keep it the way it is.
-Olof
_______________________________________________
Linuxppc-dev mailing list
Linuxppc-dev@ozlabs.org
https://ozlabs.org/mailman/listinfo/linuxppc-dev
