On (Mon) 13 Jul 2015 [10:55:55], Michael S. Tsirkin wrote:
> On Mon, Jul 13, 2015 at 01:04:45PM +0530, Amit Shah wrote:
> > That also saves us the trouble with live migration: if we hold onto a
> > buffer, that becomes state, and we'll have to migrate it as well.
> > This current model saves us from doing that.
>
> Timers hold state and have to be migrated, this seems to be missing
> here.
We do invoke virtio_rng_process in the load path, so that outstanding
requests are serviced.
Also, after migration, the guest will fetch data from the new host, so
the new host will start doing the rate-limit. We can afford to not
carry over the rate-limiting state we had from the src host.
Amit
