Re: [PATCH] regmap: irq: Make IRQ type support optional

[Charles Keepax <ckeepax@opensource.cirrus.com>]

On Fri, Dec 28, 2018 at 12:16:17PM +0000, Vaittinen, Matti wrote:
> Hello Charles,
> 
> Sending this mail form outlook web interface - so I won't inline any code :/
> 
> From: Charles Keepax [ckeepax@opensource.cirrus.com]
> Sent: Friday, December 28, 2018 1:55 PM
> 
> > On Fri, Dec 28, 2018 at 11:23:58AM +0000, Charles Keepax wrote:
> > > Currently only gpio-max77620 is using the type support in regmap IRQ,
> > > but the implementation causes the irq_set_type operation to fail on all
> > > other regmap IRQ chips. Avoid these regressions by skipping the type
> > > handling on any chips that don't define a set of supported types.
> > >
> > > Fixes: 1c2928e3e321 ("regmap: regmap-irq/gpio-max77620: add level-irq support")
> > > Signed-off-by: Charles Keepax <ckeepax@opensource.cirrus.com>
> > > ---
> > >  drivers/base/regmap/regmap-irq.c | 3 +++
> > >  1 file changed, 3 insertions(+)
> > >
> > > diff --git a/drivers/base/regmap/regmap-irq.c b/drivers/base/regmap/regmap-irq.c
> > > index 1bd1145ad8b5e..8c674f1ad0fc8 100644
> > > --- a/drivers/base/regmap/regmap-irq.c
> > > +++ b/drivers/base/regmap/regmap-irq.c
> > > @@ -257,6 +257,9 @@ static int regmap_irq_set_type(struct irq_data *data, unsigned int type)
> > >       int reg;
> > >       const struct regmap_irq_type *t = &irq_data->type;
> > >
> > > +     if (!t->types_supported)
> > > +             return 0;
> > > +
> > >       if ((t->types_supported & type) != type)
> > >               return -ENOTSUPP;
> > >
> 
> I got the bug-report from Geert and sent this patch yesterday:
> https://lore.kernel.org/lkml/20181227084443.GA23991@localhost.localdomain/
> 
> Looking at these two options, I wonder if we shuld return
> -ENOTSUPP if we do support type setting, but for example
> only for edge, not level active IRQs - and if LEVEL_LOW or
> LEVEL_HIGH is requested? Well, I have no strong opinion and
> both of these should fix the regressions - sorry for the hassle!
> 
> I still wonder whether we should do as I suggested and only
> set the irq_set_type callback for chips which have non zero
> type_registers? I suggested that here:
> https://lore.kernel.org/lkml/20181228080533.GC2461@localhost.localdomain/
> 

Yeah that does look like probably the best solution.

Thanks,
Charles