On Mon, 2016-03-21 at 12:53 +0000, Ian Campbell wrote:
>
> I hacked around this with an unsolicited CREDIT_UPDATE op for each
> buffer chain processed, but I think this is something which needs
> looking into at at the protocol level. I couldn't see anything in the
> Linux implementation which looked like it would address this, but
> perhaps I've missed something.
I finally found virtio_transport_send_credit_update() called at the end
of virtio_transport_stream_do_dequeue(), which is right about where I
would have expected it to be so I'm not sure how I missed it the first
N times I looked for it :-/.
So perhaps all which is needed here is a documentation update, since
all it says about CREDIT_UPDATE right now is that it is the reply to
CREDIT_REQUEST, not that it should be sent at the end of processing as
well.
Ian.
