of_match_table on non-DT builds

of_match_table on non-DT builds

[Rob Herring]

Grant,

With the of_match_table pointer only enabled with CONFIG_OF in struct 
device_driver, all drivers will have to have #ifdefs around their 
of_match_table if you want to build them for non-devicetree platforms. 
This is just the tip of the iceberg as all probe related OF calls have 
to be ifdef'ed. Is this the intention?

Rob

Re: of_match_table on non-DT builds

[Grant Likely]

On Tue, Nov 9, 2010 at 8:27 AM, Rob Herring <robherring2-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org> wrote:
> Grant,
>
> With the of_match_table pointer only enabled with CONFIG_OF in struct
> device_driver, all drivers will have to have #ifdefs around their
> of_match_table if you want to build them for non-devicetree platforms. This
> is just the tip of the iceberg as all probe related OF calls have to be
> ifdef'ed. Is this the intention?

Not really, and I'm not entirely sure about the best way to handle
this.  I might handle the of_match_table pointer using a macro so that
the #ifdefs can be dropped in each individual device driver
definition, or I might ask Greg to drop the #ifdefs around
of_match_table in struct device_driver, but I don't really want to
increase the size of struct device_driver for non-devicetree users.

Regardless, device drivers will still need #ifdefs around the
CONFIG_OF specific code in the probe routine; probably by breaking the
CONFIG_OF specific stuff into a helper function so that the #ifdef
isn't in the middle of a function.

g.

Re: of_match_table on non-DT builds

[Greg KH]

On Tue, Nov 09, 2010 at 09:01:17AM -0700, Grant Likely wrote:
> On Tue, Nov 9, 2010 at 8:27 AM, Rob Herring <robherring2-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org> wrote:
> > Grant,
> >
> > With the of_match_table pointer only enabled with CONFIG_OF in struct
> > device_driver, all drivers will have to have #ifdefs around their
> > of_match_table if you want to build them for non-devicetree platforms. This
> > is just the tip of the iceberg as all probe related OF calls have to be
> > ifdef'ed. Is this the intention?
> 
> Not really, and I'm not entirely sure about the best way to handle
> this.  I might handle the of_match_table pointer using a macro so that
> the #ifdefs can be dropped in each individual device driver
> definition, or I might ask Greg to drop the #ifdefs around
> of_match_table in struct device_driver, but I don't really want to
> increase the size of struct device_driver for non-devicetree users.

That's ok, it's not that big of a deal to drop the #ifdef, especially if
it makes the .c files more "sane".

I'll work on merging the dma stuff in 'struct device' to a single
pointer, to get the space back, so you should be fine.

thanks,

greg k-h