refactor(drivers/net): Unify MTUs and fix buffer sizes for virtio-net

[Gelbpunkt]

See "5.1.6.3.1 Driver Requirements: Setting Up Receive Buffers" in the VirtIO specification. We only need to allocate at least the MTU if GUEST_TSO4, GUEST_TSO6 or GUEST_UFO were negotiated, otherwise we SHOULD allocate at least 1514 (1526 minus 12 bytes header) for the payload. Currently we were allocating 65550 (MTU) and always receiving 1514 in that case.

[cagatay-y]

I think this branch will cause an error when we have MTU negotiated.

If the driver negotiates VIRTIO_NET_F_MTU, it MUST supply enough receive buffers to receive at least one receive packet of size mtu (plus low level ethernet header length) with gso_type NONE or ECN.

I don't know what MTU the device reports if guest segmentation offload is negotiated, but I think the safe behaviour is to use a constant 65550 (not the mtu) as the buffer size when it is negotiated, and if not, to use the MTU.

(VIRTIO Specification v1.2, § 5.1.4.2)

[cagatay-y]

By the way, the way we handle MTUs in general is wrong. At the point the RxQueues are created, the feature negotiation is not complete and we don't know if the device supports MTU reporting, but we try to read that field here. I have to rework the driver initialisation to avoid that.

[Gelbpunkt]

By the way, the way we handle MTUs in general is wrong. At the point the RxQueues are created, the feature negotiation is not complete and we don't know if the device supports MTU reporting, but we try to read that field here. I have to rework the driver initialisation to avoid that.

I hope I got it right now. I refactored the driver initialization to only initialize the virtqueues after the feature negotiation is complete. It probably needs some more polishing, but it looks right now. I realized that if we actually use the negotiated features, I can see that MRG_RXBUF is negotiated for me, which explains what I initially observed.

[mkroening]

Is the first committee specification draft of VIRTIO 1.3 helpful?

Virtual I/O Device (VIRTIO) Version 1.3—5.1.6.3 Setting Up Receive Buffers

So what about this?

let mut min_buf_size = our_special_mtu;

if !VIRTIO_NET_F_MRG_RXBUF {
    if any!(VIRTIO_NET_F_GUEST_TSO4, VIRTIO_NET_F_GUEST_TSO6, VIRTIO_NET_F_GUEST_UFO, VIRTIO_NET_F_GUEST_USO4, VIRTIO_NET_F_GUEST_USO6) {
        min_buf_size = usize::max(min_buf_size, 65550);
    } else {
        min_buf_size = usize::max(min_buf_size, 1514);
    }
}

if VIRTIO_NET_F_MRG_RXBUF {
    // Potentially increase min_buf_size
    // Maybe just use min_buf_size for now?
}