[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-devel] [Xenhackthon] Virtualized APIC registers - virtual interrupt delivery.

On Thu, Jun 06, 2013 at 03:08:07AM +0000, Zhang, Yang Z wrote:
> Stefano Stabellini wrote on 2013-06-05:
> > On Wed, 5 Jun 2013, Zhang, Yang Z wrote:
> >>>> Or just split the mechanism. Meaning use the event callback for "legacy"
> >>>> type events, and for PCI passthrough devices (where the host supports
> >>>> posted interrupts) just use the baremetal implementation.
> >>>> 
> >>>> That would entail some form of hypercall to identify whether a PCIe 
> >>>> device
> >>>> is "posted-interrupt" candidate and if so don't use the event channel
> >>>> mechanism for it.
> >> What you mean "a PCIe device is posted-interrupt candidate"? Do you mean
> > pass-through device will use event channel currently?
> > 
> > On Linux, yes: Linux is going to remap the MSI/MSI-X onto an event
> > channel. Therefore if we want to use posted-interrupts with pass-through
> > devices we would need to disable event channel remapping for them.
> Then we must disable event channel remapping if posted interrupt is used.

That was my understanding as well. Or at least have some form of event
channel for inter-domain communication (netback/netfront,
blkback/blkfront) - which should still be present.

Then for PCIe (so MSI and MSI-X) and IPI do not use the event channel
mechanism and just use the normal HVM type ack system.

Any ETA on when patches for this would surface?
> Best regards,
> Yang

Xen-devel mailing list



Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.