Re: I2C and devicetrees

[Grant Likely <grant.likely-s3s/WqlpOiPyB63q8FvJNQ@public.gmane.org>]

On Tue, 11 Dec 2012 13:56:08 -0700, Stephen Warren <swarren-3lzwWm7+Weoh9ZMKESR00Q@public.gmane.org> wrote:
> On 12/11/2012 01:21 PM, Olof Johansson wrote:
> > [+devicetree-discuss]
> > 
> > On Wed, Dec 5, 2012 at 2:36 PM, Peter Huewe <PeterHuewe-Mmb7MZpHnFY@public.gmane.org> wrote:
> >> Hi,
> >>
> >> I have a short question about the relations between i2c and devicetrees.
> >>
> >> I was wondering
> >> is the device part of the compatible string of a (trivial) i2c device
> >> instanciated via devicetree _always_ identical to name in i2c_client.name ?
> >> Or can it be somehow different?
> > 
> > It can be different, but the driver will then need to add a OF table
> > that matches the probing. By default the i2c/dt core code will strip
> > off the vendor prefix (before ",") and try probing with the rest of
> > the device name. If that doesn't match the client name, that is when
> > you need the additional table.
> 
> While relying on the prefix stripping works, I think I recall Grant
> mentioning that people shouldn't rely on it - namely that any I2C device
> that gets instantiated from DT should contain the OF match table
> explicitly. I CC'd Grant in case I'm mis-quoting him.

I'm not quite that strict about it. The behaviour described above is
based merely on a heuristic but for a lot of drivers that don't do
anything special it works just fine. The moment you need to identify a
specific device it becomes better to use an of match table.

g.

-- 
Grant Likely, B.Sc, P.Eng.
Secret Lab Technologies, Ltd.