Hi Tomi,

On Mon, Dec 26, 2022 at 09:25:34PM +0200, Tomi Valkeinen wrote:

On 26/12/2022 18:56, Laurent Pinchart wrote:

quoted

On Wed, Dec 14, 2022 at 08:36:47AM +0200, Tomi Valkeinen wrote:

quoted

On 14/12/2022 08:29, Tomi Valkeinen wrote:

quoted

quoted

wondering if the struct device of the DS90UB913 could be passed instead
of the port, to avoid passing the port throught
ds90ub9xx_platform_data.

Interesting thought. That would limit the number of remote i2c busses to
one, though. Not a problem for FPD-Link, but I wonder if that's assuming
too much for the future users. Then again, this is an in-kernel API so
we could extend it later if needed. So I'll try this out and see if I
hit any issues.

Right, so the issue with this one would be that it would prevent a
single device uses. E.g. a single chip which acts as an ATR (similar to
i2c-mux chips), i.e. it contains both the main and the remote i2c busses.

I don't think I understand this, sorry.

What you are suggesting above means that we'd have a separate device for 
each port of the ATR. Which is fine in our current case, as the i2c 
master busses are behind separate remote devices.

But if you consider a case similar to i2c-mux, where we have a single 
chip with the slave bus and, say, 4 master busses. We would probably 
have only a single device for that.

Hmmm... Yes you're right, it won't work in that case. Maybe we could
have two functions, the existing i2c_atr_add_adapter(), and another one
that wraps it ? It would be nice if we could get rid of the platform
data for the UB913 and UB953 drivers.

-- 
Regards,

Laurent Pinchart