On Thu, Jan 14, 2021 at 02:42:26PM +0100, Christophe Leroy wrote:
> Le 14/01/2021 à 14:22, Mark Brown a écrit :

> > For GPIO chipselects you should really fix the driver to just hand the
> > GPIO off to the core rather than trying to implement this itself, that
> > will avoid driver specific differences like this.

> IIUC, it is not trivial as it requires implementing transfer_one() instead
> of the existing transfer_one_message() in the driver. Am I right ?

Yes, that's a good idea in general though.  It should normally be pretty
simple since the conversion is mostly just deleting code doing things
which will be handled by the core.

> What's the difference/benefit of transfer_one() compared to the existing transfer_one_message() ?

It factors out all the handling of chip selects, including per-transfer
chip select inversion and so on, and also factors out all the handling
of in-message delays.  If nothing else it reduces the amount of
duplicated code that might require maintainance can have issues with
misaligned expectations from the core or client drivers.