Received-SPF: pass (google.com: domain of linux-wireless-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Message-ID: <d383ef715a7186cb382ff9450878747b8384dd0f.camel@sipsolutions.net>
Subject: Re: [PATCH] mac80211: report per-chain signal values through
 ethtool.
From:   Johannes Berg <johannes@sipsolutions.net>
To:     Ben Greear <greearb@candelatech.com>,
        linux-wireless@vger.kernel.org
Date:   Fri, 01 Jul 2022 15:08:16 +0200
In-Reply-To: <c7ee84a4-2be5-4176-2701-8c2e691a85fe@candelatech.com>
References: <20220329210228.8137-1-greearb@candelatech.com>
         <8542c3eac729563fef1bc78d28c740453fba88bd.camel@sipsolutions.net>
         <c7ee84a4-2be5-4176-2701-8c2e691a85fe@candelatech.com>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
User-Agent: Evolution 3.44.2 (3.44.2-1.fc36) 
MIME-Version: 1.0
Precedence: bulk

On Fri, 2022-07-01 at 06:05 -0700, Ben Greear wrote:
> On 7/1/22 2:55 AM, Johannes Berg wrote:
> > On Tue, 2022-03-29 at 14:02 -0700, greearb@candelatech.com wrote:
> > > From: Ben Greear <greearb@candelatech.com>
> > >=20
> > > Combine them into a u64, each byte is one chain.
> >=20
> > This only works up to 4 chains, but the specs at least support 8. I
> > don't think we have any drivers for that, but ...
>=20
> u64 gives 8 bytes, so the ethtool part can support 8 chains.
> The mac80211 part only supports up to 4 chains currently though.

Oops, right, sorry.

Still, I'm not sure I like munging it all up into one value - the value
itself then doesn't mean anything, and the normal ethtool APIs userspace
tools would be pretty much useless for it?

> > We're reporting these through nl80211 anyway though, no? Why should we
> > prefer ethtool, which fundamentally limits to a single value for the AP
> > rather than giving the full per-station view.
>=20
> I already gather ethtool stats for STA vdevs, so adding another stat is
> basically free as far as performance is concerned.  That is important
> to me.  I do not currently program much against netlink API (just scrape
> existing tools output).

Well I guess then I think you probably should program against the
netlink API :)

> > > Re-work the way that APs averaged stats to be more
> > > efficient.
> >=20
> > Isn't that completely unrelated?
>=20
> At least somewhat unrelated.
>=20

Fair enough. Maybe send that separately? I guess that's something I'd
understand a bit more and improving the existing code is an easier sell
than adding a whole new thing there :)

johannes