Re: [PATCH v4 bpf-next 02/10] ice: allow toggling loopback mode via ndo_set_features callback

[Maciej Fijalkowski <maciej.fijalkowski@intel.com>]

On Fri, Jun 17, 2022 at 06:57:41PM -0700, John Fastabend wrote:
> Maciej Fijalkowski wrote:
> > Add support for NETIF_F_LOOPBACK. This feature can be set via:
> > $ ethtool -K eth0 loopback <on|off>
> > 
> > Feature can be useful for local data path tests.
> > 
> > CC: Alexandr Lobakin <alexandr.lobakin@intel.com>
> > Signed-off-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
> > ---
> 
> Patch looks fine one question about that ice_set_features() function
> though.
> 
> Acked-by: John Fastabend <john.fastabend@gmail.com>
> 
> [...]
> 
> > +/**
> > + * ice_set_loopback - turn on/off loopback mode on underlying PF
> > + * @vsi: ptr to VSI
> > + * @ena: flag to indicate the on/off setting
> > + */
> > +static int
> > +ice_set_loopback(struct ice_vsi *vsi, bool ena)
> > +{
> > +	bool if_running = netif_running(vsi->netdev);
> > +	int ret;
> > +
> > +	if (if_running && !test_and_set_bit(ICE_VSI_DOWN, vsi->state)) {
> > +		ret = ice_down(vsi);
> > +		if (ret) {
> > +			netdev_err(vsi->netdev, "Preparing device to toggle loopback failed\n");
> > +			return ret;
> > +		}
> > +	}
> > +	ret = ice_aq_set_mac_loopback(&vsi->back->hw, ena, NULL);
> > +	if (ret)
> > +		netdev_err(vsi->netdev, "Failed to toggle loopback state\n");
> > +	if (if_running)
> > +		ret = ice_up(vsi);
> > +
> > +	return ret;
> > +}
> > +
> >  /**
> >   * ice_set_features - set the netdev feature flags
> >   * @netdev: ptr to the netdev being adjusted
> > @@ -5960,7 +5988,10 @@ ice_set_features(struct net_device *netdev, netdev_features_t features)
> >  		      clear_bit(ICE_FLAG_CLS_FLOWER, pf->flags);
> >  	}
> >  
> > -	return 0;
> > +	if (changed & NETIF_F_LOOPBACK)
> > +		ret = ice_set_loopback(vsi, !!(features & NETIF_F_LOOPBACK));
> > +
> > +	return ret;
> 
> Unrelated to your patch, but because you are messing with 'ret' here a bit,
> how come you return 0 when ice_is_safe_mode() shouldn't you push that
> error up so the user who is doing the setting knows it didn't actually
> work?

Safe mode is the first thing checked in ice_set_features() and in case it
is set we give the message to user about it and return 0 immediately.

So did you miss the immediate exit or are you suggesting that for safe
mode we should return some error code, not 0 which is interpreted as
'successful' command execution?

> 
> >  }
> >  
> >  /**
> > -- 
> > 2.27.0
> > 
> 
>