Re: [Qemu-devel] [RFC v2 3/9] virtio: stop virtqueue processing if device is broken

[Cornelia Huck <cornelia.huck@de.ibm.com>]

On Tue, 29 Mar 2016 17:12:55 +0100
Stefan Hajnoczi <stefanha@redhat.com> wrote:

> QEMU prints an error message and exits when the device enters an invalid
> state.  Terminating the process is heavy-handed.  The guest may still be
> able to function even if there is a bug in a virtio guest driver.
> 
> Moreover, exiting is a bug in nested virtualization where a nested guest
> could DoS other nested guests by killing a pass-through virtio device.
> I don't think this configuration is possible today but it is likely in
> the future.
> 
> If the broken flag is set, do not process virtqueues or write back used
> descriptors.  The broken flag can be cleared again by resetting the
> device.
> 
> Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
> ---
>  hw/virtio/virtio.c         | 39 +++++++++++++++++++++++++++++++++++++++
>  include/hw/virtio/virtio.h |  3 +++
>  2 files changed, 42 insertions(+)

> +void GCC_FMT_ATTR(2, 3) virtio_error(VirtIODevice *vdev, const char *fmt, ...)
> +{
> +    va_list ap;
> +
> +    va_start(ap, fmt);
> +    error_vreport(fmt, ap);
> +    va_end(ap);
> +
> +    vdev->broken = true;
> +
> +    if (virtio_vdev_has_feature(vdev, VIRTIO_F_VERSION_1)) {
> +        vdev->status |= VIRTIO_CONFIG_S_NEEDS_RESET;

virtio_set_status()?

> +        virtio_notify_config(vdev);
> +    }
> +}
> +
>  static void virtio_device_realize(DeviceState *dev, Error **errp)
>  {
>      VirtIODevice *vdev = VIRTIO_DEVICE(dev);
> diff --git a/include/hw/virtio/virtio.h b/include/hw/virtio/virtio.h
> index 2b5b248..1565e53 100644
> --- a/include/hw/virtio/virtio.h
> +++ b/include/hw/virtio/virtio.h
> @@ -87,6 +87,7 @@ struct VirtIODevice
>      VirtQueue *vq;
>      uint16_t device_id;
>      bool vm_running;
> +    bool broken; /* device in invalid state, needs reset */

I'm wondering whether there's a sane way to track the broken state via
the NEEDS_RESET status bit instead. We'd probably want to filter out
this bit and not expose it to legacy drivers; but as the status field
is migrated anyway, we might be able to avoid a subsection for
migration.