Re: [PATCH 09/10] events/x86: Define SCI virtual interrupt

[Boris Ostrovsky <boris.ostrovsky@oracle.com>]

On 11/07/2016 11:05 AM, Konrad Rzeszutek Wilk wrote:
> On Sun, Nov 06, 2016 at 04:42:42PM -0500, Boris Ostrovsky wrote:
>> PVH guests do not have IOAPIC which typically generates an SCI. For
>> those guests SCI will be provided as a virtual interrupt.
>>
>> We also move VIRQ_MCA definition out of xen-mca.h to
>> keep all x86-specific VIRQ_ARCH_* in one place.
> I somehow assumed that the ACPI MADT would be in the PVH guest so the
> SCI would be enumerated in there (the INT_SVC_OVR) along with the magic
> way one can figure this out.
>
> But this implies that VIRQ17 is the SCI one? But how does
> that connect to the SCI being at 9 (or 20)?
>
> I know we don't have the LAPIC in there but going forward we may really
> want it. Could you kindly explain how this VIRQ ends transumuting itself
> in the MADT value? Or would there be an event(virq)->emulated GSI 
> to be set?
>
> Sorry for not following that.

The IRQ is whatever FADT specifies (in libacpi/static_tables.c), the
guest shouldn't care. The expectation is that the guest will them bind
VIRQ_SCI to that IRQ.

So in Linux this will be something like

    if (xen_pvh_domain())
               bind_virq_to_irq(VIRQ_SCI, 0,
acpi_gbl_FADT.sci_interrupt, 0);

(and then bind_virq_to_irq() is modified to take IRQ number as argument,
if available).


>
> Also you may want to expand the PVH markdown as well to document how the
> CPU hotplug mechanism will work (or where you thinking to do that once
> you get feedback on this, in which case pls ignore me).


I'll add a comment there.

-boris



_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
https://lists.xen.org/xen-devel