Re: [PATCH net-next v2 0/8] ethtool: add pause frame stats

[Vladimir Oltean <olteanv@gmail.com>]

On Fri, Sep 11, 2020 at 05:07:24PM -0700, Jakub Kicinski wrote:
> On Sat, 12 Sep 2020 02:49:32 +0300 Vladimir Oltean wrote:
> > On Fri, Sep 11, 2020 at 04:28:45PM -0700, Jakub Kicinski wrote:
> > > Hi!
> > >
> > > This is the first (small) series which exposes some stats via
> > > the corresponding ethtool interface. Here (thanks to the
> > > excitability of netlink) we expose pause frame stats via
> > > the same interfaces as ethtool -a / -A.
> > >
> > > In particular the following stats from the standard:
> > >  - 30.3.4.2 aPAUSEMACCtrlFramesTransmitted
> > >  - 30.3.4.3 aPAUSEMACCtrlFramesReceived
> > >
> > > 4 real drivers are converted, hopefully the semantics match
> > > the standard.
> > >
> > > v2:
> > >  - netdevsim: add missing static
> > >  - bnxt: fix sparse warning
> > >  - mlx5: address Saeed's comments
> >
> > DSA used to override the "ethtool -S" callback of the host port, and
> > append its own CPU port counters to that.
> >
> > So you could actually see pause frames transmitted by the host port and
> > received by the switch's CPU port:
> >
> > # ethtool -S eno2 | grep pause
> > MAC rx valid pause frames: 1339603152
> > MAC tx valid pause frames: 0
> > p04_rx_pause: 0
> > p04_tx_pause: 1339603152
> >
> > With this new command what's the plan?
>
> Sounds like something for DSA folks to decide :)
>
> What does ethtool -A $cpu_port control?
> The stats should match what the interface controls.

Error: $cpu_port: undefined variable.
With DSA switches, the CPU port is a physical Ethernet port mostly like
any other, except that its orientation is inwards towards the system
rather than outwards. So there is no network interface registered for
it, since I/O from the network stack would have to literally loop back
into the system to fulfill the request of sending a packet to that
interface.
The ethtool -S framework was nice because you could append to the
counters of the master interface while not losing them.
As for "ethtool -A", those parameters are fixed as part of the
fixed-link device tree node corresponding to the CPU port.