On Fri, Mar 13, 2026 at 9:37 AM Di Zhu [off-list ref] wrote:

quoted

Although VIRTIO_NET_F_CTRL_GUEST_OFFLOADS is negotiated, which
indicates the device supports dynamic control of guest offloads, it
does not necessarily mean the device supports specific hardware GRO features.

If none of the features defined in GUEST_OFFLOAD_GRO_HW_MASK (such as
TSO4, TSO6, or UFO) are present in vi->guest_offloads_capable, the
device effectively lacks the hardware capability to perform GRO.

In this case, NETIF_F_GRO_HW should be cleared from dev->hw_features
to prevent the stack from attempting to enable an unsupported hardware
offload configuration.

Fixes: a02e8964eaf9 ("virtio-net: ethtool configurable LRO")
Signed-off-by: Di Zhu <redacted>
---
 drivers/net/virtio_net.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c index

72d6a9c6a5a2..49a60af684d5 100644

--- a/drivers/net/virtio_net.c
+++ b/drivers/net/virtio_net.c

@@ -7058,6 +7058,9 @@ static int virtnet_probe(struct virtio_device *vdev)
        }
        vi->guest_offloads_capable = vi->guest_offloads;

+       if (!(vi->guest_offloads_capable & GUEST_OFFLOAD_GRO_HW_MASK))
+               dev->hw_features &= ~NETIF_F_GRO_HW;
+

It looks to me the root cause is this during probe?

        if (virtio_has_feature(vdev, VIRTIO_NET_F_CTRL_GUEST_OFFLOADS))
                dev->hw_features |= NETIF_F_GRO_HW;

Thanks

	Exactly. It seems the code blindly enables NETIF_F_GRO_HW as long as the 
control channel is present, without verifying if the backend actually supports the
offload features.

quoted

        rtnl_unlock();

        err = virtnet_cpu_notif_add(vi);
--
2.34.1