On 2020/2/27 上午1:48, Yuri Benditovich wrote:
> Support for VIRTIO_NET_F_RSS feature in QEMU for reference
> purpose. Implements Toeplitz hash calculation for incoming
> packets according to configuration provided by driver.
>
> This series requires previously submitted and accepted
> patch to be applied:
> https://lists.gnu.org/archive/html/qemu-devel/2020-01/msg06448.html
>
> Yuri Benditovich (3):
> virtio-net: introduce RSS RX steering feature
> virtio-net: implement RSS configuration command
> virtio-net: implement RX RSS processing
>
> hw/net/trace-events | 3 +
> hw/net/virtio-net.c | 234 +++++++++++++++++++-VIRTIO_NET_F_RSS
> include/hw/virtio/virtio-net.h | 12 +
> include/standard-headers/linux/virtio_net.h | 37 +++-
> 4 files changed, 273 insertions(+), 13 deletions(-)
>
One question before the reviewing.
Do we need to deal with migration (which I think yes)?
I think we don't (yet) as this is a reference implementation and the main goal is to provide the functional reference for hardware solution.
I agree with the general direction that for complete support of RSS and hash delivery the best way is to do that in kernel using bpf.
For that, IMO, the bpf should be a part of the kernel (it uses skb fields) and the tap should receive just RSS parameters for it.
At this stage we definitely will need to add migration support and propagation of RSS parameters.
Thanks