Re: [PATCH v1 net-next] net: dsa: qca: ar9331: add ethtool stats support

[Jakub Kicinski]

On Fri, 20 Nov 2020 13:59:07 +0100 Oleksij Rempel wrote:
> On Mon, Nov 16, 2020 at 04:28:44PM -0800, Jakub Kicinski wrote:
> > On Tue, 17 Nov 2020 02:10:05 +0200 Vladimir Oltean wrote:  
> > > On Mon, Nov 16, 2020 at 04:02:13PM -0800, Jakub Kicinski wrote:  
> > > > For a while now we have been pushing back on stats which have a proper
> > > > interface to be added to ethtool -S. So I'd expect the list of stats
> > > > exposed via ethtool will end up being shorter than in this patch.    
> > > 
> > > Hmm, not sure if that's ever going to be the case. Even with drivers
> > > that are going to expose standardized forms of counters, I'm not sure
> > > it's going to be nice to remove them from ethtool -S.  
> > 
> > Not remove, but also not accept adding them to new drivers.
> >   
> > > Testing teams all
> > > over the world have scripts that grep for those. Unfortunately I think
> > > ethtool -S will always remain a dumping ground of hell, and the place
> > > where you search for a counter based on its name from the hardware block
> > > guide as opposed to its standardized name/function. And that might mean
> > > there's no reason to not accept Oleksij's patch right away. Even if he
> > > might volunteer to actually follow up with a patch where he exposes the
> > > .ndo_get_stats64 from DSA towards drivers, as well as implements
> > > .ndo_has_offload_stats and .ndo_get_offload_stats within DSA, that will
> > > most likely be done as separate patches to this one, and not change in
> > > any way how this patch looks.  
> 
> Ok, so what is the plan for me? Implement .ndo_get_stats64 before this
> one?

Yup. And ethtool_ops::get_pause_stats, preferable 'cause I think you
had pause stats there, too.