Re: [PATCH] gpio: mxs: use dynamic allocation of base

[Alexander Stein <alexander.stein@ew.tq-group.com>]

Hi everyone,

Am Dienstag, 28. Februar 2023, 09:46:27 CET schrieb Marco Felsch:
> Hi Dario,
> 
> On 23-02-28, Dario Binacchi wrote:
> > Hi Stefan,
> > 
> > On Mon, Feb 27, 2023 at 6:52 PM Stefan Wahren <stefan.wahren@i2se.com> 
wrote:
> > > Hi Dario,
> > > 
> > > Am 26.02.23 um 21:53 schrieb Dario Binacchi:
> > > > Since commit 502df79b860563d7 ("gpiolib: Warn on drivers still using
> > > > static
> > > > gpiobase allocation"), one or more warnings are printed during boot on
> > > > systems where static allocation of GPIO base is used:
> > > > 
> > > > [    0.136834] gpio gpiochip0: Static allocation of GPIO base is
> > > > deprecated, use dynamic allocation. [    0.142753] gpio gpiochip1:
> > > > Static allocation of GPIO base is deprecated, use dynamic allocation.
> > > > [    0.148452] gpio gpiochip2: Static allocation of GPIO base is
> > > > deprecated, use dynamic allocation. [    0.154341] gpio gpiochip3:
> > > > Static allocation of GPIO base is deprecated, use dynamic allocation.
> > > > [    0.160097] gpio gpiochip4: Static allocation of GPIO base is
> > > > deprecated, use dynamic allocation.
> > > > 
> > > > So let's follow the suggestion and use dynamic allocation.
> > > 
> > > i understand the motivation of avoiding these warnings.
> > > 
> > > AFAIK the MXS platform is older than the dynamic allocation of GPIO
> > > base. In the perfect world all applications has been migrated to
> > > libgpiod / chardev GPIO. But i'm really concerned this hasn't happend
> > > yet, at least for this platform. So i believe this change break
> > > applications, since it affects userspace.
> > 
> > Thanks for your explanations.
> 
> There was also a lengthly discussion here:
> 
> https://lore.kernel.org/all/20230120104647.nwki4silrtd7bt3w@pengutronix.de/
> 
> > Do you think it makes sense to use a Kconfig option to enable/disable
> > the dynamic allocation of GPIO base?
> 
> What we could do is to check the GPIO_SYSFS kconfig symbol. If this is
> enabled we should keep the current behaviour to not break the uAPI else
> we can go with the dynamic numbering. What we should avoid is a new
> Kconfig symbol.

Even that is only half of the truth. Apparently this is crucial for 
GPIO_SYSFS, but even without sysfs it affects the order of the character 
devices, which might break userspace. The main open issue is, whether GPIO 
aliases shall be supported or not, see see [1] and [2].
Once this is clear, the switch to dynamic base can be done.

Best regards,
Alexander

[1] https://lore.kernel.org/linux-gpio/20230215092421.143199-1-alexander.stein@ew.tq-group.com/T/#u
[2] https://lore.kernel.org/linux-gpio/5941337.lOV4Wx5bFT@steina-w/T/
#m2a74777fc4031b7f6f30ac1dcf65b20e7d3f8f12

> Regards,
>   Marco
> 
> > As done, if I'm not mistaken, in commit 80d34260f36c6 ("pinctrl:
> > renesas: gpio: Use dynamic GPIO base if no function GPIOs")?
> > With legacy support enabled by default, but the ability to try out the
> > latest dynamic allocation functionality.
> > 
> > Thanks and regards,
> > Dario
> > 
> > > Best regards
> > > 
> > > > Signed-off-by: Dario Binacchi <dario.binacchi@amarulasolutions.com>
> > > > ---
> > > > 
> > > >   drivers/gpio/gpio-mxs.c | 2 +-
> > > >   1 file changed, 1 insertion(+), 1 deletion(-)
> > > > 
> > > > diff --git a/drivers/gpio/gpio-mxs.c b/drivers/gpio/gpio-mxs.c
> > > > index 7f59e5d936c2..b48a7c1fb7c1 100644
> > > > --- a/drivers/gpio/gpio-mxs.c
> > > > +++ b/drivers/gpio/gpio-mxs.c
> > > > @@ -330,7 +330,7 @@ static int mxs_gpio_probe(struct platform_device
> > > > *pdev)
> > > > 
> > > >       port->gc.to_irq = mxs_gpio_to_irq;
> > > >       port->gc.get_direction = mxs_gpio_get_direction;
> > > > 
> > > > -     port->gc.base = port->id * 32;
> > > > +     port->gc.base = -1;
> > > > 
> > > >       err = gpiochip_add_data(&port->gc, port);
> > > >       if (err)


-- 
TQ-Systems GmbH | Mühlstraße 2, Gut Delling | 82229 Seefeld, Germany
Amtsgericht München, HRB 105018
Geschäftsführer: Detlef Schneider, Rüdiger Stahl, Stefan Schneider
http://www.tq-group.com/