Thanks for posting this as a separate patch, it makes the review easier.
The idea and the patch look good to me.
One comment:
with this code we don't need the ctxs array, we can just have a single
ctx in the for loop.
Other than that:
Acked-by: Daniele Di Proietto
Packet batch sorting is removed for three reasons:
1) The following patches for NAT change the locking
marshalling so batching loses benefit.
2) For real mixtures of flows either in hypervisors
or gateways, the batch sorting won't provide benefit
and will just be a tax.
3) Code