On Fri, Nov 28, 2014 at 08:48:24AM +0008, Jason Wang wrote:

On Fri, Nov 28, 2014 at 4:10 AM, Michael S. Tsirkin [off-list ref] wrote:

quoted

The spec states that mac in config space is only driver-writable in the
legacy case.  Fence writing it in virtnet_set_mac_address() in the
virtio 1.0 case.

Suggested-by: Cornelia Huck <redacted>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
---
drivers/net/virtio_net.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
index c6a72d3..9ab3c50 100644
--- a/drivers/net/virtio_net.c
+++ b/drivers/net/virtio_net.c

@@ -1030,7 +1030,8 @@ static int virtnet_set_mac_address(struct net_device

*dev, void *p)
				 "Failed to set mac address by vq command.\n");
			return -EINVAL;
		}
-	} else if (virtio_has_feature(vdev, VIRTIO_NET_F_MAC)) {
+	} else if (virtio_has_feature(vdev, VIRTIO_NET_F_MAC) &&
+		   !virtio_has_feature(vdev, VIRTIO_F_VERSION_1)) {
		unsigned int i;
		/* Naturally, this has an atomicity problem. */
-- 
MST

Does this mean there's no way to setting mac addres
if ctrl mac addr is disabled in virtio-1?

More exactly: no way for host to know the guest mac address.

If yes, we'd better keep this compatibility somehow.

If host wants to be notified about guest mac updates,
it must enable the appropriate ctrl command.
Or if it's an old host, it just doesn't enable virtio-1.

There's no way around this - it's in the spec.


-- 
MST