[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <CA+sq2CcCSX4NuoyZZ020xH75s1+ii5Xc6XM5jAovUTpQtaZXfQ@mail.gmail.com>
Date: Mon, 13 Jan 2020 16:57:06 +0530
From: Sunil Kovvuri <sunil.kovvuri@...il.com>
To: Jakub Kicinski <kubakici@...pl>
Cc: Linux Netdev List <netdev@...r.kernel.org>,
"David S. Miller" <davem@...emloft.net>,
Christina Jacob <cjacob@...vell.com>,
Sunil Goutham <sgoutham@...vell.com>
Subject: Re: [PATCH 14/17] octeontx2-pf: Add basic ethtool support
On Sat, Jan 11, 2020 at 6:57 PM Jakub Kicinski <kubakici@...pl> wrote:
>
> On Sat, 11 Jan 2020 14:17:45 +0530, Sunil Kovvuri wrote:
> > On Sat, Jan 11, 2020 at 12:58 AM Jakub Kicinski wrote:
> > > On Fri, 10 Jan 2020 17:11:58 +0530, sunil.kovvuri@...il.com wrote:
> > > > +static const struct otx2_stat otx2_dev_stats[] = {
> > > > + OTX2_DEV_STAT(rx_bytes),
> > > > + OTX2_DEV_STAT(rx_frames),
> > > > + OTX2_DEV_STAT(rx_ucast_frames),
> > > > + OTX2_DEV_STAT(rx_bcast_frames),
> > > > + OTX2_DEV_STAT(rx_mcast_frames),
> > > > + OTX2_DEV_STAT(rx_drops),
> > > > +
> > > > + OTX2_DEV_STAT(tx_bytes),
> > > > + OTX2_DEV_STAT(tx_frames),
> > > > + OTX2_DEV_STAT(tx_ucast_frames),
> > > > + OTX2_DEV_STAT(tx_bcast_frames),
> > > > + OTX2_DEV_STAT(tx_mcast_frames),
> > > > + OTX2_DEV_STAT(tx_drops),
> > > > +};
> > >
> > > Please don't duplicate the same exact stats which are exposed via
> > > ndo_get_stats64 via ethtool.
> >
> > ndo_stats64 doesn't have separate stats for ucast, mcast and bcast on Rx and
> > Tx sides, they are combined ones. Hence added separate stats here.
> > The ones repeated here are bytes, frames and drops which are added to have
> > full set of stats at one place which could help anyone debugging pkt
> > drop etc issues.
>
> Same exact as in bytes, frames, and drops are exactly the same rather
> than e.g. one being counted by hardware and the other by software.
>
> No objection to reporting the *cast stats broken out via ethtool.
I am not getting what your objection here is, what's reported via stats64 and
ethtool both are counted by hardware. I have checked other NIC drivers and they
also do report overall pkts, bytes, drop counters via ethtool.
Is there any harm in reporting this way ?
Thanks,
Sunil.
Powered by blists - more mailing lists