Re: Re: [PATCH v12 09/11] gpio: pl061: switch data path to gpio-regmap

[Andy Shevchenko <andriy.shevchenko@linux.intel.com>]

On Tue, Oct 27, 2026 at 06:32:04PM +0800, zl020895 wrote:
> At 2026-10-27 16:06:15, "Andy Shevchenko" <andriy.shevchenko@linux.intel.com> wrote:
> >On Tue, Oct 27, 2026 at 12:18:56PM +0800, Long Zhao via B4 Relay wrote:

...

> >> +	if (irq) {
> >> +		irq_set_chained_handler_and_data(irq, pl061_irq_handler, pl061);
> >
> >> +		pl061->irq_handler_installed = true;
> >
> >Can't we remove the chained handler unconditionally?
> >
> >> +	}

> On the chained-handler teardown: pl061_remove_irq() is registered
> before devm_gpio_regmap_register(), so on unbind it runs after
> gpiochip_remove(). gpiochip_free_remaining_irqs() therefore still
> sees the irq domain and its mappings; the action then disposes
> them. The chained handler is installed only after
> devm_gpio_regmap_register() succeeds. If that call fails, the
> action still runs although no handler was installed, which is
> why irq_handler_installed exists.
> 
> Options for v13:
> A) Keep irq_handler_installed and document this ordering.
> B) Split into two devm actions. The early one, still registered
>    before the gpiochip, only calls irq_dispose_mapping(), so
>    unbind order is unchanged. A second action, registered right
>    after irq_set_chained_handler_and_data(), clears the handler
>    unconditionally, and the flag goes away.
> 
> Do you prefer A or B?

Option B seems plausible.

-- 
With Best Regards,
Andy Shevchenko