On Thu, Mar 16, 2017 at 10:22:27AM +0100, Maxime Coquelin wrote:
>
>
> On 03/16/2017 08:35 AM, Yuanhan Liu wrote:
> >On Tue, Mar 14, 2017 at 12:02:59PM +0100, Maxime Coquelin wrote:
> >>>+uint64_t
> >>>+rte_vhost_get_negotiated_features(int vid)
> >>>+{
> >>>+ struct virtio_net *dev;
> >>>+
> >>>+ dev = get_device(vid);
> >>>+ if (!dev)
> >>>+ return -1;
> >>>+
> >>>+ return dev->features;
> >>>+}
> >>Are we sure the negotiation is done when we can get the device?
> >
> >Yes. However, one thing worth noting is that the features may change
> >after the new_device() callback. Notablely, when live migration starts
> >/ends, the VHOST_F_LOG_ALL will be set/cleared, respectively.
>
> Good point.
>
> >From that point of view, we need a new callback, something like
> >features_changed(), or live_migration_starts()? Or a better name?
>
> I would name it features_changed() to be future-proof (even if I don't
> foresee other cases than live migration right now).
Right, it will be "features_changed()" then.
--yliu
