Re: [libvirt] [PATCH v2 10/25] qemu: capabilities: Add virtio/vhost {non-}transitional

On Tue, 2019-01-29 at 15:48 +0100, Ján Tomko wrote: > On Wed, Jan 23, 2019 at 04:32:38PM -0500, Cole Robinson wrote: >> This adds QEMU_CAPS flags for the following devices >> >> virtio-blk-pci-transitional >> virtio-blk-pci-non-transitional >> virtio-net-pci-transitional >> virtio-net-pci-non-transitional >> vhost-scsi-pci-transitional >> vhost-scsi-pci-non-transitional >> virtio-rng-pci-transitional >> virtio-rng-pci-non-transitional >> virtio-9p-pci-transitional >> virtio-9p-pci-non-transitional >> virtio-balloon-pci-transitional >> virtio-balloon-pci-non-transitional >> vhost-vsock-pci-transitional >> vhost-vsock-pci-non-transitional >> virtio-input-host-pci-transitional >> virtio-input-host-pci-non-transitional >> virtio-scsi-pci-transitional >> virtio-scsi-pci-non-transitional >> virtio-serial-pci-transitional >> virtio-serial-pci-non-transitional > > This seems excessive, is there a plan to retire the transitional > devices? I don't expect anyone creating a QEMU build that e.g.: > a) supports virtio-rng-pci-transitional but not virtio-rng-pci-non-transitional > b) supports virtio-scsi-pci-transitional but not virtio-input-host-pci-transitional > > For the disable-legacy property, we only have > QEMU_CAPS_VIRTIO_PCI_DISABLE_LEGACY, that is checked on multiple (but > not all possible) PCI devices. That's a very good point! We could have a single capability QEMU_CAPS_VIRTIO_PCI_NON_TRANSITIONAL that is set if any out of a bunch of {,non-}transitional devices is present, and key everything else off that... Eduardo, do you think we might ever get in trouble if we did that? For example, because of QEMU dropping transitional devices but leaving non-transitional devices in?