Hi Mike On 10/26/25 7:29 PM, Michael Kelly wrote:
> To perform the translation by the ACPI server it is necessary to pass -1 > for the irq number and supply the PCI parameters and these are not > readily available to the code at that point. Even if it were possible to > find those parameters I don't think it would change the size of the code > change required compared to doing the translation in ddekit_pci_irq_enable. Yes, you're right, sorry. > It would indeed be quite simple to call the acpi_server directly rather > than using libirqhelp. My thinking was that that library was intended to > abstract all IRQ behaviour (and implementation details) in one place > that would avoid callers needing to know how the IRQ was derived. I've > no strong opinion either way, really. I think it makes sense to make an RPC directly to acpi server, because it was not expected that clients of irqhelp would require the irq before the handler was registered and lose the pci b/d/f information. Usually at the point of knowing the b/d/f for the device, we also register the handler, (for example in the pci-userspace code of rump). Eager to see your patches landing. Damien
