On 19/04/2018 17:59, Michael S. Tsirkin wrote: > On Thu, Apr 19, 2018 at 05:51:51PM +0200, Paolo Bonzini wrote: >> On 19/04/2018 17:19, Michael S. Tsirkin wrote: >>>> - if we make it 1 when weak barriers are needed, the device also needs >>>> to nack feature negotiation (not allow setting the FEATURES_OK) if the >>>> bit is not set by the driver. >>>> However, that is not enough. Live >>>> migration assumes that it is okay to migrate a virtual machine from a >>>> source that doesn't support a feature to a destination that supports it. >>>> In this case, it would assume that it is okay to migrate from software >>>> virtio to hardware virtio. This is wrong because the destination would >>>> use weak barriers >>> >>> You can't migrate between systems with different sets of device features >>> right now. >> >> Yes, you can, exactly because some features are defined not by the >> machine type but rather by the host kernel. See virtio_net_get_features >> in QEMU's hw/virtio/virtio-net.c, and virtio_set_features_nocheck in >> QEMU's hw/virtio/virtio.c. > > Oh you are right. Well we can just special-case that one :)
Anything wrong with bumping the PCI revision? Paolo