> Hi Ilya,
> 
> Please find my replies inline.

Hi all, this patch seems to have hit a wall, I can see there is a v6 that came 
after this that has been acked: 
http://patchwork.ozlabs.org/project/openvswitch/patch/[email protected]/

I would like to see this work go in, but think we should finish the discussion 
on this tread before proceeding. My thoughts below on the discussion.

<snip> 

> > > I have already made a POC based on the Discussion above and will put
> > > that patch in few days with the 'dp_extra_info' showing DPCLS as well.
> > >  As for this Patch is also important as currently the get-command doesn’t
> > show much info and hence this improvement will make things clear for the
> > user through the dpcl-get-command.
> > > Hence request this usability feature to be merged.
> > > I will follow this patch with the 'dp_extra_info' showing DPCLS functions
> > next to give absolute clarity to the user.
> >
> > Why the 'subtable-lookup-prio-get' should return anything except for the
> > priority of subtable lookup implementations?
> >
> 
> We think the DPIF, MFEX and DPCLS commands should all provide the same
> workflow to the user.
> By having "subtable-lookup-prio-get" return the list (including "use count" 
> item)
> allows the get/set
> pair of commands to both modify and check the results of setting the DPCLS
> implementations.
> This patch makes the output and workflow the same as DPIF and MFEX.
> 

I agree with the above sentiment. After testing for the past few days with the 
various DPIF/MEF functions it seems more natural to me to have the required 
info here in the dpcls command.

I  tried the approach in the POC posted by Amber to provide the info in the 
flow dump and while this approach could also have the stats info, I felt there 
was a disconnect in the sense why would this info not be accessible via a dpcls 
command?

http://patchwork.ozlabs.org/project/openvswitch/patch/[email protected]/

In fairness to the folks from both RH and Intel here who have worked on this 
over the past few months, reviewing the previous revisions, I can see the 
reasoning for providing this info in this dpcls command and would agree with 
the approach so would like to see this approach go ahead.

> > That doesn't seem straightforward.  Especially if we'll have that 
> > information
> > reported in a different place already without need for extra infrastructure.
> >
> 
> The "dump-flows" approach provides information between the flow and the
> subtable implementaiton
> which is also useful, however it does not provide the above workflow like DPIF
> and MFEX. We are willing
> to also accept the "dump-flows" approach, but the "subtable lookup prio get"
> patch/workflow is required.

+1, we could also include this approach as a separate merge if that makes 
things more amenable, it might mean some duplication but it would be a fair 
compromise for end user intuitiveness.

Thanks
Ian

> 
> Regards
> Amber
> 
> > >
> > > Regards
> > > Amber
> > >> Best regards, Ilya Maximets.
> 

_______________________________________________
dev mailing list
[email protected]
https://mail.openvswitch.org/mailman/listinfo/ovs-dev

Reply via email to