Re: [PATCH] gpio: siox: potentially enabling IRQs too early

[Thomas Gleixner <tglx@xxxxxxxxxxxxx>]

Uwe,

Uwe Kleine-König <u.kleine-koenig@xxxxxxxxxxxxxx> writes:
> On Tue, Feb 11, 2020 at 09:59:30AM +0100, Thomas Gleixner wrote:
>> Uwe Kleine-König <u.kleine-koenig@xxxxxxxxxxxxxx> writes:
>> Indeed. Looking at the driver, all of the spin_lock_irq() usage in the
>> irqchip callbacks is broken.
>> 
>> So this needs two changes:
>> 
>>    1) Convert to raw spin lock, as this won't work on RT otherwise
>> 
>>    2) s/lock_irq/lock/ for all irqchip callbacks.
>
> Are you sure about the calls in gpio_siox_get_data()? This is only
> called by siox_poll() which doesn't disable irqs.

I explicitely said: "for all irqchip callbacks".

gpio_siox_get_data() is not a irq chip callback, right? So obviously it
has to stay there.

Thanks,

        tglx