Re: [net-next PATCH 3/4] net: sched: cls_u32 add bit to specify software only rules

[Jamal Hadi Salim <jhs@mojatatu.com>]

On 16-02-24 11:04 PM, John Fastabend wrote:
> On 16-02-24 05:31 AM, Jamal Hadi Salim wrote:

> I think this is absolutely necessary not only for performance of
> reporting the rules back to software but if we don't do it generically
> the driver will have to do it anyways because doing the inverse
> transformation from hw implementation to u32 is really tricky and in
> fact with hnodes and knodes there are multiple cls_u32 "programs" that
> functionally are the same so we have no way to return what the user
> actually programmed without it. Further eBPF (the next classifier I'm
> working on) is even worse in this regard.

Ok, I guess there are multiple use cases for it ;->
Yes, with ebpf it will be worse because data and instructions are
inter- mingled (and our interest is in data only). Note:
Over the years this has been a big struggle for human
friendliness. I thought we didnt care about humans (as in automation)
but you are saying this will affect machines too ;-> We cant allow
that ;->

Note: You could decode u32 descriptions but it is an involved effort.
Example, see this feature in tc:
---
jhs@jhs1 tc -pretty filter ls dev $ETH parent ffff: protocol ip
filter pref 11 u32
filter pref 11 u32 fh 800: ht divisor 1
filter pref 11 u32 fh 800::800 order 2048 key ht 800 bkt 0 flowid 1:1
   match IP src 10.0.0.130/32
	action order 1: gact action drop
	 random type none pass val 0
	 index 1 ref 1 bind 1
----
See that "match" field reading in anglais? It requires more and more
additions of pretty printers that translate back.

What about adding some tag to allow for easy "babel translation"?

> You can see my solution to
> this "load in hardware" filter list in patch 4/4. See Jiri's comment
> also on that and see if you agree.

Ok, will do.

cheers,
jamal