On 09/09/2016 07:48 PM, Michael S. Tsirkin wrote:
On Fri, Sep 09, 2016 at 03:10:07PM +0200, Maxime Coquelin wrote:This patch makes pci devices plugging more robust, by not confusing guest with modern interface when the backend doesn't support VIRTIO_F_VERSION_1. Cc: Cornelia Huck <cornelia.h...@de.ibm.com> Cc: Marcel Apfelbaum <mar...@redhat.com> Cc: Michael S. Tsirkin <m...@redhat.com> Cc: qemu-sta...@nongnu.org Signed-off-by: Maxime Coquelin <maxime.coque...@redhat.com> --- hw/virtio/virtio-pci.c | 15 +++++++++++++++ hw/virtio/virtio-pci.h | 5 +++++ 2 files changed, 20 insertions(+) diff --git a/hw/virtio/virtio-pci.c b/hw/virtio/virtio-pci.c index 755f921..9e88d7b 100644 --- a/hw/virtio/virtio-pci.c +++ b/hw/virtio/virtio-pci.c @@ -1581,6 +1581,21 @@ static void virtio_pci_device_plugged(DeviceState *d, Error **errp) uint32_t size; VirtIODevice *vdev = virtio_bus_get_device(&proxy->bus); + /* + * Virtio capabilities present without + * VIRTIO_F_VERSION_1 confuses guests + */ + if (!virtio_test_backend_virtio_1(vdev, errp)) { + virtio_pci_disable_modern(proxy); + } + + legacy = virtio_pci_legacy(proxy); + modern = virtio_pci_modern(proxy); + if (!legacy && !modern) { + error_setg(errp, "PCI device is neither legacy nor modern."); + return; + } +How does this interact with virtio-pci: error out when both legacy and modern modes are disabled ? If it's the same, I'd rather pick that one and apply your change on top.
Not exactly, since with my patch modern can be disabled after realize callback has been called. But I will base my next revision on top of Greg's patch, and use the same pattern as he did use. Thanks, Maxime
