This series makes device plugging more robust, to avoid guest to be confused
when the backend doesn't support VIRTIO_F_VERSION_1.
The problem is seen with Linux guests running mainline kernels, when backend
doesn't support the feature:
virtio_net virtio0: virtio: device uses modern interface but does not have
VIRTIO_F_VERSION_1.
When it happens, the modern device probe returns -EINVAL, whereas its caller
expects -ENODEV being returned to switch to legacy device probing.
We need to make QEMU more robust to ensure the guest won't be confused,
so this series exposes modern interface only when backend support it.
It has been tested with vhost-net and vhost-user backends in client
and server modes.
Changes since v1:
-----------------
- Make the backend feature check function specialized to only VIRTIO_1
Maxime Coquelin (2):
virtio: Add function to check whether backend supports VIRTIO_1
virtio-pci: Disable modern interface if backend without
VIRTIO_F_VERSION_1
hw/virtio/virtio-pci.c | 15 +++++++++++++++
hw/virtio/virtio-pci.h | 5 +++++
hw/virtio/virtio.c | 13 +++++++++++++
include/hw/virtio/virtio.h | 1 +
4 files changed, 34 insertions(+)
--
2.7.4
