Re: [Qemu-devel] [PATCH 0/4] Tweaks around virtio-blk start/stop

[Paolo Bonzini <pbonzini@redhat.com>]

On 16/03/2016 13:22, Cornelia Huck wrote:
>> > Yeah, it doesn't help that the functions are underdocumented (as in the
>> > "assign" parameter above).
> My understanding is:
> 
> - 'assign': set up a new notifier (true) or disable it (false)
> - 'set_handler': use our handler (true) or have it handled elsewhere
>   (false)

Right.  So if we're setting up a new notifier in
virtio_queue_aio_set_host_notifier_handler, virtio_pci_stop_ioeventfd should

> > > I don't think the ->set_host_notifiers() api really allows for this.
> > 
> > I think it does, assign is the last argument to k->set_host_notifier().
> 
> This depends on whether we want 'assign' to clean up any old notifiers
> before setting up new ones. I think we want different behaviour for
> dataplane and vhost.

I think dataplane and vhost are the same.

The question is whether ioeventfd=off,vhost=on or
ioeventfd=off,dataplane=on are valid combinations; I think they aren't.

If they aren't, it should be okay to remove the
virtio_queue_host_notifier_read call in
virtio_queue_set_host_notifier_fd_handler and
virtio_queue_aio_set_host_notifier_handler.  That's because a handler
for the notifier will always be set _somewhere_.  It could be the usual
ioeventfd handler, the vhost handler or the dataplane handler, but one
will be there.

Paolo