Hi,


As you suggested I will split the drivers. I am thinking of doing 5 patches:

- factor out master() parts

- separate Master part to i2c-designware-master.c (changes in i2c-designware-core.c)

- enable Slave part to i2c-designware-slave (changes in i2c-designware-core.c)

- glue drivers and device bindings

- cleaning


Regards,

Luis

On 21-Oct-16 11:54, Andy Shevchenko wrote:

On Fri, 2016-10-21 at 10:56 +0100, Luis Oliveira wrote:

quoted

Since practically 90% of the code is shared between master and slave,
I was
thinking if it will be acceptable to use the same driver for both but
differentiate the master/slave mode by the compatible strings.

It might be possible to split like other drivers do:

1. Core part (i2c-designware-core.c)
2. Master part (i2c-designware-master.c)
3. Slave part (i2c-designware-slave.c)
4. Glue drivers (like: i2c-designware-platdrv.c)

quoted

Thanks,
Luis

On 10/18/2016 16:17, Wolfram Sang wrote:

quoted

quoted

This is needed because the configuration is different and the i2c-
designware
cannot be master/slave without a reset. To resolve that I added
this property
to bind it as a slave when needed.

Aww, pity that the HW can't do that. Do you have details why?

If that is really a HW limitation, then I'd suggest having a
seperate
driver for slave-only mode so we can differentiate by compatible
strings.