On Wed, Jun 20, 2018 at 11:04 AM, Robert Jarzmik robert.jarzmik@free.fr wrote:
Rob Herring robh+dt@kernel.org writes:
On Wed, Jun 20, 2018 at 8:52 AM, Robert Jarzmik robert.jarzmik@free.fr wrote:
Rob Herring robh+dt@kernel.org writes:
Hi Rob,
wm9713@0 {audio-codec@0
Sure.
reg = <0>; /* Codex index (between 0 and 3) */ compatible = "ac97-codec";This should be something with 'wm9713' in it or if there are ID registers you can base it on that.
Well, let's discuss that. If I take as an example sdio, in mmc-card.txt, the compatible string is "mmc-card", which describes the "kind" of subnode it is, not _exactly_ the subnode it is.
"mmc-card" is only for memory cards which are pretty much all the same and rarely have a node, but SDIO devices always have a specific compatible.
Does AC97 specify a common programming model? That's were a common compatible is used (though experience has proven things like "generic-ehci" are don't work).
It's generally a bad design to mix different types of subnodes, so I was assuming all child nodes would be the same type. For example, you can only have 1 address space (i.e. reg definition). If you need different types, you need to add a level of nodes.
If the purpose for this compatible string is to filter the child nodes which are only ac97 codecs, so that the ac97 framework can act upon them (ie. acquire the clock for example), then wm9713 is too specific ... That won't work for another device with an Analog Devices AD1835 for example.
Are there ID registers? If so, then just construct a compatible string from them and just find the compatible child node. See USB bindings for an example using VID/PID.
Rob