Re: [PATCH net-next v9 2/5] net: spacemit: Add K1 Ethernet MAC

[Simon Horman <horms@kernel.org>]

On Tue, Sep 09, 2025 at 05:15:56PM +0100, Simon Horman wrote:
> On Sat, Sep 06, 2025 at 12:35:37AM +0800, Vivian Wang wrote:
> > On 9/6/25 00:01, Simon Horman wrote:
> > 
> > > On Fri, Sep 05, 2025 at 11:45:29PM +0800, Vivian Wang wrote:
> > >
> > > ...
> > >
> > > Hi Vivian,
> > >
> > >>>> +		status = emac_rx_frame_status(priv, rx_desc);
> > >>>> +		if (unlikely(status == RX_FRAME_DISCARD)) {
> > >>>> +			ndev->stats.rx_dropped++;
> > >>> As per the comment in struct net-device,
> > >>> ndev->stats should not be used in modern drivers.
> > >>>
> > >>> Probably you want to implement NETDEV_PCPU_STAT_TSTATS.
> > >>>
> > >>> Sorry for not mentioning this in an earlier review of
> > >>> stats in this driver.
> > >>>
> > >> On a closer look, these counters in ndev->stats seems to be redundant
> > >> with the hardware-tracked statistics, so maybe I should just not bother
> > >> with updating ndev->stats. Does that make sense?
> > > For rx/tx packets/bytes I think that makes sense.
> > > But what about rx/tx drops?
> > 
> > Right... but tstats doesn't have *_dropped. It seems that tx_dropped and
> > rx_dropped are considered "slow path" for real devices. It makes sense
> > to me that those should be very rare.
> > 
> > So it seems that what I should do is to just track tx_dropped and
> > rx_dropped myself in a member in emac_priv and report in the
> > ndo_get_stats64 callback, and use the hardware stuff for the rest, as
> > implemented now.
> 
> Thanks, that makes sense to me.

Oops, for the 2nd time today I see I've responded where other's have
already done so. In this case, please take Jakub's advice elsewhere
in this thread.