Re: [PATCH 4/5] checks: check for #{size,address}-cells without child nodes

[David Gibson <david-xT8FGy+AXnRB3Ne2BGzF6laj5H9X9Tb+@public.gmane.org>]

[-- Attachment #1: Type: text/plain, Size: 2217 bytes --]

On Mon, Nov 20, 2017 at 10:22:53AM -0600, Rob Herring wrote:
> On Sun, Nov 19, 2017 at 6:14 PM, David Gibson
> <david-xT8FGy+AXnRB3Ne2BGzF6laj5H9X9Tb+@public.gmane.org> wrote:
> > On Fri, Nov 17, 2017 at 08:45:14AM -0600, Rob Herring wrote:
> >> nodes with a 'reg' property nor a ranges property.
> >>
> >> An exception may be an overlay that adds nodes, but this case would
> >> need
> >
> > Sentence doesn't seem finished..
> 
> It was there until I rebased. Since the line started with #, git dropped it...
> 
> "#{size,address}-cells in the overlay to properly compile already."

Ah, right.

> > In any case, I'm not sure this is a good idea.  It's not uncommon to
> > have bus bridge nodes that ought to have a well defined #address and
> > #size cells, but just don't happen to have any plugged devices yet.
> > An overlay that adds nodes is one possibility, but a bus where the
> > children can be probed is another.
> 
> Wouldn't an overlay without #{size,address}-cells have warnings from
> avoid_default_addr_size?

Well, yes.  The checks are pretty much designed for whole trees and
don't work well on overlays at the moment.  That's a rather more
substantial change to fix.

> As long as there are no child nodes, the check is not run.

Ah!  Sorry, I missed that.  Objection withdrawn in that case.

> So we're
> limited to false positives if we have a mixture of nodes with and
> without unit addresses and only the nodes without unit addresses are
> populated. I have seen this with PCI hosts with an interrupt
> controller child node.

Sounds like an incorrect representation if the intrrupt controller
doesn't have a PCI config space presence.

> In general, I'm struggling with how to have tests that check for
> things that we generally want to avoid, but we still allow exceptions.
> Some of these may be things we want to avoid in new bindings, but
> wouldn't fix for existing cases. Another example is things that
> were/are valid for OF, but not FDT.

-- 
David Gibson			| I'll have my music baroque, and my code
david AT gibson.dropbear.id.au	| minimalist, thank you.  NOT _the_ _other_
				| _way_ _around_!
http://www.ozlabs.org/~dgibson

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]