cvs commit: src/sys/dev/pci pci.c

[Marcel Moolenaar]

On Fri, Dec 03, 2004 at 01:26:50PM -0700, M. Warner Losh wrote:
> : > We likely should have a sys/$MACHINE_ARCH/include/_config.h that has a
> : > bunch of #defines that we can test instead of specific architecture
> : > lists, don't you think?
> : 
> : A <machine/_config.h> doesn't give me warm fuzzies. Maybe we can
> : rename <machine/pci_cfgreg.h> to <machine/pci.h> and give it some
> : additional use (for this particular case of course).
> 
> There's a growing number of these sorts of things.  NetBSD has this
> information in machine/types.h.  For example:
> 
> #include <mips/types.h>
> 
> #define __HAVE_GENERIC_SOFT_INTERRUPTS
> #define __HAVE_DEVICE_REGISTER
> #define __HAVE_NWSCONS
> 
> #define __HAVE_MIPS_MACHDEP_CACHE_CONFIG
> 
> I was thinking that we should start doing something similar...

For some reason I have a bad feeling about this. It's soooo easily
misused. For one, it's a good way to cover up a bad design or an
incomplete implementation. It goes no further than a feeling though.
There's nothing wrong with it in principle, because differences do
exist.

-- 
 Marcel Moolenaar	  USPA: A-39004		 marcel at xcllnt.net