Re: [PATCH] ssb: Fix DMA-API compilation for non-PCI systems

[Andrew Morton <akpm@linux-foundation.org>]

On Thu, 6 Nov 2008 22:23:09 +0100
Michael Buesch <mb@bu3sch.de> wrote:

> On Thursday 06 November 2008 22:16:00 Mike Frysinger wrote:
> > On Thu, Nov 6, 2008 at 15:49, Michael Buesch wrote:
> > > --- wireless-testing.orig/include/linux/ssb/ssb.h       2008-08-01 17:26:05.000000000 +0200
> > > +++ wireless-testing/include/linux/ssb/ssb.h    2008-11-06 21:45:37.000000000 +0100
> > > @@ -427,12 +427,16 @@ static inline int ssb_dma_mapping_error(
> > >  {
> > >        switch (dev->bus->bustype) {
> > >        case SSB_BUSTYPE_PCI:
> > > +#ifdef CONFIG_SSB_PCIHOST
> > >                return pci_dma_mapping_error(dev->bus->host_pci, addr);
> > > +#endif
> > > +               break;
> > >        case SSB_BUSTYPE_SSB:
> > >                return dma_mapping_error(dev->dev, addr);
> > >        default:
> > > -               __ssb_dma_not_implemented(dev);
> > > +               break;
> > >        }
> > 
> > all these functions now read:
> > default: break;
> > seems kind of pointless ... why not just drop that case completely
> 
> Because the compiler complains "not handled all cases...".
> And yes, we do want to trigger __ssb_dma_not_implemented() for
> these cases.

Please always quote the compiler error messages when fixing build
errors.  It is unobvious what the problems are here.  I could struggle
away and create a CONFIG_PCI=n build, but what .c file do I need to
compile?  Dunno.


All those ifdefs are nasty.  Couldn't we do something like:

#ifdef CONFIG_SSB_PCIHOST
static inline int
ssb_pci_dma_mapping_error(structy pci_dev *host_pci, dma_addr_t addr)
{
	return pci_dma_mapping_error(host_pci, addr);
}

#else

static inline int
ssb_pci_dma_mapping_error(structy pci_dev *host_pci, dma_addr_t addr)
{
	return -ENOSYS;
}

#endif

(etc)

and then leave the __ssb_dma_not_implemented() calls under the default:
case in the switch statements?