Re: [PATCH] init. mca_bus_type even if !MCA_bus

[Greg KH <greg@kroah.com>]

On Mon, May 17, 2004 at 03:52:22PM -0700, Andrew Morton wrote:
> "Randy.Dunlap" <rddunlap@osdl.org> wrote:
> >
> > On Mon, 17 May 2004 15:14:12 -0700 Andrew Morton wrote:
> > 
> > | "Randy.Dunlap" <rddunlap@osdl.org> wrote:
> > | >
> > | > -	if(mca_system_init()) {
> > | > +	if (mca_system_init()) {
> > | >  		printk(KERN_ERR "MCA bus system initialisation failed\n");
> > | >  		return -ENODEV;
> > | >  	}
> > | >  
> > | > +	if (!MCA_bus)
> > | > +		return -ENODEV;
> > | 
> > | Why is it appropriate to register the MCA bus type when there is no
> > | MCA bus present?
> > 
> > Mostly because it was selected with CONFIG_MCA=y.
> > 
> > Another option (I think, need to test) is to check !MCA_bus
> > in drivers/mca/mca-legacy.c::find_mca_adapter(), so that
> > mca_bus_type isn't used when it shouldn't be.
> > 
> > Do you prefer that approach?
> 
> well my question really was a question, rather than a reimplementation
> suggestion.  If it _is_ appropriate that mca_bus_type be registered on a
> platform which is discovered to have no MCA hardware then fine.
> 
> Greg? James?  Any insights?

Ick, James is the one to answer here, as I remember this being pretty
messy at times due to the way the MCA code currently is...

thanks,

greg k-h