Re: [PATCH v3 2/2] s390/pci: Migrate s390 IRQ logic to IRQ domain API

[Thomas Gleixner]

On Tue, Nov 18 2025 at 17:13, Tobias Schumacher wrote:

>  static struct irq_chip zpci_irq_chip = {
>  	.name = "PCI-MSI",
>  	.irq_unmask = pci_msi_unmask_irq,
>  	.irq_mask = pci_msi_mask_irq,
> +	.irq_compose_msi_msg = zpci_compose_msi_msg
>  };

> +static struct msi_parent_ops zpci_msi_parent_ops = {
> +	.supported_flags   = MSI_GENERIC_FLAGS_MASK	|
> +			     MSI_FLAG_PCI_MSIX		|
> +			     MSI_FLAG_MULTI_PCI_MSI,
> +	.required_flags	   = MSI_FLAG_USE_DEF_DOM_OPS  |
> +			     MSI_FLAG_USE_DEF_CHIP_OPS |
> +			     MSI_FLAG_PCI_MSI_MASK_PARENT,

That MASK_PARENT flag is really only necessary if you want to avoid
masking/unmasking at the PCI level during operation
(disable/enable_irq()). See

f09c1d63e895 ("irqchip/msi-lib: Honor the MSI_FLAG_PCI_MSI_MASK_PARENT flag")

for a detailed explanation.

But as s390 does not seem to provide mask/unmask at a different level of
the interrupt transport, setting this flag and the mask/unmask callbacks
above is pointless.

If the flag is not set the PCI core will use pci_msi_[un]mask_irq() for the per
device chip at the top of the hierarchy, which avoids the indirection to
the parent chip.

Thanks,

        tglx