Re: [Qemu-devel] Using aio_poll for timer carrier threads

[Stefan Hajnoczi]

On Wed, Aug 14, 2013 at 11:05:29AM +0200, Jan Kiszka wrote:
> On 2013-08-14 10:52, Stefan Hajnoczi wrote:
> > On Tue, Aug 13, 2013 at 04:13:03PM +0200, Jan Kiszka wrote:
> >> On 2013-08-13 15:45, Stefan Hajnoczi wrote:
> >>> On Tue, Aug 13, 2013 at 09:56:17AM +0200, Jan Kiszka wrote:
> >>> The details depend on your device, do you have a git repo I can look at
> >>> to understand your device model?
> >>
> >> Pushed my hacks here:
> >>
> >> git://git.kiszka.org/qemu.git queues/rt.new3
> > 
> > Excellent, thanks!  Are you calling qemu_raise_irq() outside the global
> > mutex and how is it protected?
> 
> By luck and via many exceptions, specifically by disabling of HPET
> support (to avoid that it is involved in IRQ routing - or even used in
> legacy mode) and by relying on the direct delivery to the kernel in KVM
> mode. Yes, IRQ delivery is still a huge construction site for BQL-free
> device models.

Okay.  In dataplane I use the guest notifier for virtio devices (irqfd
for KVM mode with MSI-X or bounced through
virtio_queue_guest_notifier_read()).

Stefan