Re: [PATCH 09/14] serial: sh-sci: Introduced function pointers

[Thierry Bultel <thierry.bultel.yh@bp.renesas.com>]

On Mon, Feb 10, 2025 at 03:45:02PM +0100, Geert Uytterhoeven wrote:
> Hi Thierry,
> 

Hi Geert,

> On Wed, 29 Jan 2025 at 18:03, Thierry Bultel
> <thierry.bultel.yh@bp.renesas.com> wrote:
> > The aim here is to prepare support for new sci controllers like
> > the T2H/RSCI whose registers are too much different for being
> > handled in common code.
> >
> > This named serial controller also has 32 bits register,
> > so some return types had to be changed.
> >
> > The needed generic functions are no longer static, with prototypes
> > defined in sh-sci-common.h so that they can be used from specific
> > implementation in a separate file, to keep this driver as little
> > changed as possible.
> >
> > For doing so, a set of 'ops' is added to struct sci_port.
> >
> > Signed-off-by: Thierry Bultel <thierry.bultel.yh@bp.renesas.com>
> 
> Thanks for your patch!
> 
> I can't say I am super-enthusiastic about this approach.
> The SCI variant in RZ/T2 seems to differ a lot from the already
> supported SCI, SCIF, SCIFA, SCIFB, and HSCIF variants. The latter
> are very similar, but usually have just more features/registers, and
> further differ in a few different register offsets and bit locations.
> If you compare the RZ/T2 SCI block diagram with the SH7751 SCI block
> diagram (or even the R-Car SCIF block diagram), the most striking
> similarity is that both have a baud rate generator that can divide
> Pclk by 1, 4, 16, or 64 ;-)
> So perhaps you're better off adding a completely new driver?
> 

The origin of this work is the non-mainline 'rz-sci.c' from renesas BSP,
which is a actually separate driver, but that we considered 'similar enough'
(a lot of duplicated lines) to be integrated in sh-sci with some work.

The number of places where to add specific code for the new port type 
were however a concern in terms of non-regression tests and future
evolutions.
So the idea was basically to change it once in the most possible modular
way, so that coming features and/or bug fixes keep specific to rzsci.

I am looking forward to getting more minds about that approach, which
seemed the less intrusive, keeping as much code in common as possible.

> What do other people think?
> Thanks!
> 
> Gr{oetje,eeting}s,
> 
>                         Geert
> 
> -- 
> Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org
> 
> In personal conversations with technical people, I call myself a hacker. But
> when I'm talking to journalists I just say "programmer" or something like that.
>                                 -- Linus Torvalds