Re: [PATCH] Fix readw/writew warnings in drivers/net/wireless/hermes.h

[Ben Dooks <ben@fluff.org.uk>]

On Mon, Oct 11, 2004 at 02:39:00PM +0200, Jan Dittmer wrote:
> David Gibson wrote:
> >On Mon, Oct 11, 2004 at 02:29:39PM +0200, Jan Dittmer wrote:
> >
> >>Cal Peake wrote:
> >>
> >>>On Mon, 11 Oct 2004, Jan Dittmer wrote:
> >>>
> >>>
> >>>
> >>>>Cal Peake wrote:
> >>>>
> >>>>
> >>>>
> >>>>>	inw((hw)->iobase + ( (off) << (hw)->reg_spacing )) : \
> >>>>>-	readw((hw)->iobase + ( (off) << (hw)->reg_spacing )))
> >>>>>+	readw((void __iomem *)(hw)->iobase + ( (off) << (hw)->reg_spacing )))
> >>>>>#define hermes_write_reg(hw, off, val) do { \
> >>>>
> >>>>Isn't the correct fix to declare iobase as (void __iomem *) ?
> >>>
> >>>
> >>>iobase is an unsigned long, declaring it as a void pointer is prolly not 
> >>>what we want to do here. The typecast seems proper. A lot of other 
> >>>drivers do this as well thus it must be proper ;-)
> >>
> >>Why is iobase a unsigned long in the first place? Isn't this broken for 
> >>64bit archs?
> >
> >
> >Um, no.
> >
> 
> Yeah, just rememberd when sending the mail ;-). Still, most drivers seem 
> to use (void __iomem *) in the declaration of their iobase.

IIRC, ioremap() and friends all return (void __iomem *) or at least
(void *)

-- 
Ben (ben@fluff.org, http://www.fluff.org/)

  'a smiley only costs 4 bytes'