On 21/03/14 16:13, Laurent Pinchart wrote: > Hi Tomi, > > On Friday 21 March 2014 15:37:17 Tomi Valkeinen wrote: >> On 21/03/14 00:32, Laurent Pinchart wrote: >>> The OF graph bindings documentation could just specify the ports node as >>> optional and mandate individual device bindings to specify it as mandatory >>> or forbidden (possibly with a default behaviour to avoid making all >>> device bindings too verbose). >> >> Isn't it so that if the device has one port, it can always do without >> 'ports', but if it has multiple ports, it always has to use 'ports' so >> that #address-cells and #size-cells can be defined? > > You can put the #address-cells and #size-cells property in the device node > directly without requiring a ports subnode.
Ah, right. So 'ports' is only needed when the device node has other children nodes than the ports and those nodes need different #address-cells and #size-cells than the ports. In that case it sounds fine to leave it for the driver bindings to decide. Tomi
signature.asc
Description: OpenPGP digital signature
