Re: [PATCH pinctrl REGRESSION] Revert "pinctrl: mvebu: armada-37xx: use use platform api"

[Pali Rohár <pali@xxxxxxxxxx>]

On Tuesday 07 April 2020 13:52:30 Pali Rohár wrote:
> On Tuesday 24 March 2020 13:20:17 Andrew Lunn wrote:
> > On Tue, Mar 24, 2020 at 01:44:13AM +0100, Marek Behún wrote:
> > > This reverts commit 06e26b75f5e613b400116fdb7ff6206a681ab271.
> > > 
> > > This commit caused a regression on Armada 37xx. The pinctrl driver says
> > >   armada-37xx-pinctrl d0013800.pinctrl: invalid or no IRQ
> > >   armada-37xx-pinctrl d0018800.pinctrl: invalid or no IRQ
> > > and afterwards other drivers cannot use GPIOs by this driver as IRQs.
> > > 
> > > Fixes: 06e26b75f5e6 ("pinctrl: mvebu: armada-37xx: use use platform...")
> > > Signed-off-by: Marek Behún <marek.behun@xxxxxx>
> > > Cc: Peng Fan <peng.fan@xxxxxxx>
> > > ---
> > >  drivers/pinctrl/mvebu/pinctrl-armada-37xx.c | 12 +++---------
> > >  1 file changed, 3 insertions(+), 9 deletions(-)
> > > 
> > > diff --git a/drivers/pinctrl/mvebu/pinctrl-armada-37xx.c b/drivers/pinctrl/mvebu/pinctrl-armada-37xx.c
> > > index 32f12a388b3c..5f125bd6279d 100644
> > > --- a/drivers/pinctrl/mvebu/pinctrl-armada-37xx.c
> > > +++ b/drivers/pinctrl/mvebu/pinctrl-armada-37xx.c
> > > @@ -15,6 +15,7 @@
> > >  #include <linux/of.h>
> > >  #include <linux/of_address.h>
> > >  #include <linux/of_device.h>
> > > +#include <linux/of_irq.h>
> > >  #include <linux/pinctrl/pinconf-generic.h>
> > >  #include <linux/pinctrl/pinconf.h>
> > >  #include <linux/pinctrl/pinctrl.h>
> > > @@ -741,14 +742,7 @@ static int armada_37xx_irqchip_register(struct platform_device *pdev,
> > >  		return ret;
> > >  	}
> > >  
> > > -	nr_irq_parent = platform_irq_count(pdev);
> > 
> > Hi Marek
> > 
> > Could you determine the value of nr_irq_parent(). Is it -EPROBE_DEFER?
> 
> Hello Andrew! I have tested it with 5.6 kernel and return value in
> nr_irq_parent is in both cases zero. So it is not -EPROBE_DEFER. And
> return value of of_irq_count(np) is 12 for d0013800.pinctrl and 5 for
> d0018800.pinctrl.

Adding Jason, Gregory and Sebastian into the loop.

Could you please look at this problem? If there is no easy solution,
I would suggest to revert problematic commit as pinctrl in current state
is broken and unusable.

> So this is a clear regression introduced by that patch.
> 
> > This patch has removed the handling of that.
> > 
> > > -	if (nr_irq_parent < 0) {
> > > -		if (nr_irq_parent != -EPROBE_DEFER)
> > > -			dev_err(dev, "Couldn't determine irq count: %pe\n",
> > > -				ERR_PTR(nr_irq_parent));
> > > -		return nr_irq_parent;
> > > -	}
> > 
> > Thanks
> > 	Andrew