On Mon, Mar 21, 2016 at 2:48 PM, Michael Hennerich
[off-list ref] wrote:

The thing is that the device directly supports an OPEN DRAIN mode.

If the consumer asks for open drain mode - everything works as expected.
However instead of utilizing the on-chip feature, GPIOLIB will emulate open
drain behaviour by configuring the chip for input when someone writes logic
high.

So the comment above was really meant to say -
remove on-chip OPEN DRAIN mode, consumers should use the GPIOLIB
implementation if required.

This is very interesting.

So if a hardware has *real* open drain, i.e. a bit in hardware that says
"set me in open drain mode" we should manage this in gpiolib
and provide a special callback in the vtable for it, right? Because it
is certainly better to do this with hardware, if we can.

I'll send a draft patch and we can see if this is what you want.

Yours,
Linus Walleij