Re: [PATCH v2 3/8] irqchip/qcom-pdc: Remove pdc_enable_intr() wrapper

[Maulik Shah (mkshah)]

On 5/26/2026 6:04 PM, Stephan Gerhold wrote:
> On Tue, May 26, 2026 at 04:24:39PM +0530, Maulik Shah wrote:
>> pdc->enable_intr() function already points to respective version
>> specific enable function. pdc_enable_intr() now only kept as wrapper.
>> Remove the wrapper and invoke pdc->enable_intr() from caller.
>>
>> Locking in pdc_enable_intr() applies lock to all pdc->enable_intr()
>> however its only required for pdc_enable_intr_bank() which uses
>> a shared bank across all interrupts. pdc_enable_intr_cfg() do not
>> required locking as IRQ_CFG registers are one per interrupt. Move
>> locking accordingly.
> 
> Well, pdc_enable_intr_cfg() is still a read-modify-write. If two CPUs
> read IRQ_i_CFG at the same time and modify different bits (e.g. enable
> and type bits) then write back the modified register one of the
> modifications will get lost. Can we be sure that this won't happen?

Two or more CPUs can not handle/configure same IRQ at the same time
as the irq_desc->lock would serialize all irqchip callbacks for same irq.

Thanks,
Maulik