Re: [PATCH] gpio: pca953x: Improve interrupt support

From: work work
Date: Tue Jan 14 2025 - 10:45:25 EST


Le mar. 14 janv. 2025 à 10:37, Andy Shevchenko
<andy.shevchenko@xxxxxxxxx> a écrit :
>
> On Tue, Jan 14, 2025 at 12:03 AM lakabd <lakabd.work@xxxxxxxxx> wrote:
> >
....

> > + /* Store irq_mask for later use when checking pending IRQs */
> > + bitmap_or(chip->unmasked_interrupts, chip->unmasked_interrupts, chip->irq_mask, gc->ngpio);
>
> This solution has a flaw. Where is any code that clears this new
> bitmap? The code starts with 0 (obviously) and step by step it gets
> saturated to all-1s.
>

Yes indeed, and actually the new bitmap is not necessary at all
because what we need does already exist which is chip->irq_mask (I
noticed it just now!).
chip->irq_mask is updated whenever a pin is masked or unmasked via
pca953x_irq_mask() and pca953x_irq_unmask().

The solution should look like this:

diff --git a/gpio-pca953x.c b/gpio-pca953x.c
index 272febc..29e8c20 100644
--- a/gpio-pca953x.c
+++ b/gpio-pca953x.c
@@ -842,11 +842,6 @@ static bool pca953x_irq_pending(struct
pca953x_chip *chip, unsigned long *pendin
int ret;

if (chip->driver_data & PCA_PCAL) {
- /* Read the current interrupt status from the device */
- ret = pca953x_read_regs(chip, PCAL953X_INT_STAT, trigger);
- if (ret)
- return false;
-
/* Check latched inputs and clear interrupt status */
ret = pca953x_read_regs(chip, chip->regs->input, cur_stat);
if (ret)
@@ -855,7 +850,7 @@ static bool pca953x_irq_pending(struct
pca953x_chip *chip, unsigned long *pendin
/* Apply filter for rising/falling edge selection */
bitmap_replace(new_stat, chip->irq_trig_fall, chip->irq_trig_raise,
cur_stat, gc->ngpio);

- bitmap_and(pending, new_stat, trigger, gc->ngpio);
+ bitmap_and(pending, new_stat, chip->irq_mask, gc->ngpio);

return !bitmap_empty(pending, gc->ngpio);
}

--
Best Regards
Abderrahim LAKBIR