Re: [PATCH 2/3] gpio: bcm-kona: Make sure GPIO bits are unlocked when requesting IRQ

[Florian Fainelli <florian.fainelli@broadcom.com>]

On 1/30/25 13:46, Artur Weber wrote:
> On 30.01.2025 22:35, Florian Fainelli wrote:
>> On 1/30/25 09:10, Artur Weber wrote:
>>> diff --git a/drivers/gpio/gpio-bcm-kona.c b/drivers/gpio/gpio-bcm-kona.c
>>> index 
>>> 77bd4ec93a231472d7bc40db9d5db12d20bb1611..eeaa921df6f072129dbdf1c73d6da2bd7c1fe716 100644
>>> --- a/drivers/gpio/gpio-bcm-kona.c
>>> +++ b/drivers/gpio/gpio-bcm-kona.c
>>> ...
>>> @@ -87,14 +103,25 @@ static void bcm_kona_gpio_lock_gpio(struct 
>>> bcm_kona_gpio *kona_gpio,
>>>       unsigned long flags;
>>>       int bank_id = GPIO_BANK(gpio);
>>>       int bit = GPIO_BIT(gpio);
>>> +    struct bcm_kona_gpio_bank *bank = &kona_gpio->banks[bank_id];
>>> -    raw_spin_lock_irqsave(&kona_gpio->lock, flags);
>>> +    if (bank->gpio_unlock_count[bit] == 0) {
>>> +        dev_err(kona_gpio->gpio_chip.parent,
>>> +            "Unbalanced locks for GPIO %u\n", gpio);
>>> +        return;
>>
>> Don't you want to release &kona_gpio->lock here?
>>
> 
> If you're talking about the "raw_spin_lock_irqsave" call above the if
> condition - note that it has been removed and was actually moved below:

Oh my bad, my eye corrected that - to a + somehow.

> 
>>> +    if (bank->gpio_unlock_count[bit] == 0) {
>>> +        raw_spin_lock_irqsave(&kona_gpio->lock, flags);
>>> +
>>> +        val = readl(kona_gpio->reg_base + GPIO_PWD_STATUS(bank_id));
>>> +        val |= BIT(bit);
>>> +        bcm_kona_gpio_write_lock_regs(kona_gpio->reg_base, bank_id, 
>>> val);
>>> +
>>> +        raw_spin_unlock_irqrestore(&kona_gpio->lock, flags);
>>> +    }
> 
> In general, the diff for this patch did not generate very well, sorry
> about that. I'll see if there's anything I can do to improve that in the
> future.

Nah that's fine, thanks!
-- 
Florian