Re: [PATCH v3 09/11] serial: sc16is7xx: add I/O register translation offset

[Hugo Villeneuve <hugo@xxxxxxxxxxx>]

On Thu, 25 May 2023 11:31:45 -0400
Hugo Villeneuve <hugo@xxxxxxxxxxx> wrote:

> On Thu, 25 May 2023 14:22:46 +0300
> andy.shevchenko@xxxxxxxxx wrote:
> 
> > Thu, May 25, 2023 at 12:03:23AM -0400, Hugo Villeneuve kirjoitti:
> > > From: Hugo Villeneuve <hvilleneuve@xxxxxxxxxxxx>
> > > 
> > > If the shared GPIO pins on a dual port/channel variant like the
> > > SC16IS752 are configured as GPIOs for port A, and modem control lines
> > > on port A, we need to translate the Linux GPIO offset to an offset
> > > that is compatible with the I/O registers of the SC16IS7XX (IOState,
> > > IODir and IOIntEna).
> > > 
> > > Add a new variable to store that offset and set it when we detect that
> > > special case.
> > 
> > ...
> > 
> > > +/*
> > > + * We may need to translate the Linux GPIO offset to a SC16IS7XX offset.
> > > + * This is needed only for the case where a dual port variant is configured to
> > > + * have only port B as modem status lines.
> > > + *
> > > + * Example for SC16IS752/762 with upper bank (port A) set as GPIOs, and
> > > + * lower bank (port B) set as modem status lines (special case described above):
> > > + *
> > > + * Pin         GPIO pin     Linux GPIO     SC16IS7XX
> > > + * name        function     offset         offset
> > > + * --------------------------------------------------
> > > + * GPIO7/RIA    GPIO7          3              7
> > > + * GPIO6/CDA    GPIO6          2              6
> > > + * GPIO5/DTRA   GPIO5          1              5
> > > + * GPIO4/DSRA   GPIO4          0              4
> > > + * GPIO3/RIB    RIB           N/A            N/A
> > > + * GPIO2/CDB    CDB           N/A            N/A
> > > + * GPIO1/DTRB   DTRB          N/A            N/A
> > > + * GPIO0/DSRB   DSRB          N/A            N/A
> > > + *
> > > + * Example  for SC16IS750/760 with upper bank (7..4) set as modem status lines,
> > 
> > Single space is enough.
> 
> Fixed.
> 
>  
> > > + * and lower bank (3..0) as GPIOs:
> > > + *
> > > + * Pin         GPIO pin     Linux GPIO     SC16IS7XX
> > > + * name        function     offset         offset
> > > + * --------------------------------------------------
> > > + * GPIO7/RI     RI            N/A            N/A
> > > + * GPIO6/CD     CD            N/A            N/A
> > > + * GPIO5/DTR    DTR           N/A            N/A
> > > + * GPIO4/DSR    DSR           N/A            N/A
> > > + * GPIO3        GPIO3          3              3
> > > + * GPIO2        GPIO2          2              2
> > > + * GPIO1        GPIO1          1              1
> > > + * GPIO0        GPIO0          0              0
> > > + */
> > 
> > Wondering if you can always register 8 pins and use valid mask to define which
> > one are in use?
> 
> I will look into it, I think it may be a good idea and could help to simplify things a bit.

Hi,
finally, this was the way to go. The resulting code/patch is much simpler and elegant this way. Thank you for the suggestion.

I will submit a V4 with all the changes.

Hugo.