Re: [PATCH v3 07/13] PCI: apple: Move away from INTMSK{SET,CLR} for INTx and private interrupts

[Marc Zyngier <maz@kernel.org>]

On 2025-04-22 18:41, Bjorn Helgaas wrote:
> On Tue, Apr 01, 2025 at 10:17:07AM +0100, Marc Zyngier wrote:
>> T602x seems to have dropped the rather useful SET/CLR accessors
>> to the masking register.
>> 
>> Instead, let's use the mask register directly, and wrap it with
>> a brand new spinlock. No, this isn't moving in the right direction.
> 
>> @@ -261,14 +262,16 @@ static void apple_port_irq_mask(struct irq_data 
>> *data)
>>  {
>>  	struct apple_pcie_port *port = irq_data_get_irq_chip_data(data);
>> 
>> -	writel_relaxed(BIT(data->hwirq), port->base + PORT_INTMSKSET);
>> +	guard(raw_spinlock_irqsave)(&port->lock);
>> +	rmw_set(BIT(data->hwirq), port->base + PORT_INTMSK);
> 
> sparse v0.6.4-39-gce1a6720 complains about this (and similar usage
> elsewhere):
> 
>   $ make C=2 drivers/pci/
>   drivers/pci/controller/pcie-apple.c:311:13: warning: context
> imbalance in 'apple_port_irq_mask' - wrong count at exit
>   drivers/pci/controller/pcie-apple.c:319:13: warning: context
> imbalance in 'apple_port_irq_unmask' - wrong count at exit
> 
> But I guess we just have to live with this for now until somebody
> makes sparse smarter:
> 
> https://lore.kernel.org/linux-sparse/CAHk-=wiVDZejo_1BhOaR33qb=pny7sWnYtP4JUbRTXkXCkW6jA@mail.gmail.com/
> 
> Nothing to do, just "huh".

Huh indeed. Pretty sad to see a useful piece of tooling being left
on the curb.

GSoC project?

         M.
-- 
Jazz is not dead. It just smells funny...