Re: [PATCH 1/2] gpio: em: remove the gpiochip before removing the irq domain

From: Geert Uytterhoeven
Date: Wed Jul 10 2019 - 06:13:19 EST


Hi Phil,

On Wed, Jul 10, 2019 at 11:37 AM Phil Reid <preid@xxxxxxxxxxxxxxxxx> wrote:
> On 10/07/2019 17:08, Bartosz Golaszewski wrote:
> > From: Bartosz Golaszewski <bgolaszewski@xxxxxxxxxxxx>
> >
> > In commit 8764c4ca5049 ("gpio: em: use the managed version of
> > gpiochip_add_data()") we implicitly altered the ordering of resource
> > freeing: since gpiochip_remove() calls gpiochip_irqchip_remove()
> > internally, we now can potentially use the irq_domain after it was
> > destroyed in the remove() callback (as devm resources are freed after
> > remove() has returned).
> >
> > Use devm_add_action() to keep the ordering right and entirely kill
> > the remove() callback in the driver.
> >
> > Reported-by: Geert Uytterhoeven <geert@xxxxxxxxxxxxxx>
> > Fixes: 8764c4ca5049 ("gpio: em: use the managed version of gpiochip_add_data()")
> > Cc: stable@xxxxxxxxxxxxxxx
> > Signed-off-by: Bartosz Golaszewski <bgolaszewski@xxxxxxxxxxxx>

> > --- a/drivers/gpio/gpio-em.c
> > +++ b/drivers/gpio/gpio-em.c

> > @@ -333,39 +340,32 @@ static int em_gio_probe(struct platform_device *pdev)
> > return -ENXIO;
> > }
> >
> > + ret = devm_add_action(&pdev->dev,
> > + em_gio_irq_domain_remove, p->irq_domain);
>
> Could devm_add_action_or_reset be used?

Thank you very much for bringing this function to my attention!
I was just wondering if devm_add_action() should call the action on
failure, as this is what most callers seem to do anyway.

>
> > + if (ret) {
> > + irq_domain_remove(p->irq_domain);
> > + return ret;
> > + }

Gr{oetje,eeting}s,

Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@xxxxxxxxxxxxxx

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds