Re: [PATCH net-next] bnxt_en: Allow ntuple filters for drops

[Joe Damato <joe@dama.to>]

On Wed, Jan 28, 2026 at 03:09:21PM -0800, Michael Chan wrote:
> On Wed, Jan 28, 2026 at 2:27 PM Joe Damato <joe@dama.to> wrote:
> >
> > It appears that in commit 7efd79c0e689 ("bnxt_en: Add drop action
> > support for ntuple"), bnxt gained support for ntuple filters for packet
> > drops.
> >
> > However, support for this does not seem to work in recent kernels or
> > against net-next:
> >
> >   % sudo ethtool -U eth0 flow-type udp4 src-ip 1.1.1.1 action -1
> >     rmgr: Cannot insert RX class rule: Operation not supported
> >     Cannot insert classification rule
> >
> > The issue is that the existing code uses ethtool_get_flow_spec_ring_vf,
> > which will return a non-zero value if the ring_cookie is set to
> > RX_CLS_FLOW_DISC, which then causes bnxt_add_ntuple_cls_rule to return
> > -EOPNOTSUPP because it thinks the user is trying to set an ntuple filter
> > for a vf.
> >
> > Fix this by first checking that the ring_cookie is not RX_CLS_FLOW_DISC.
> 
> Please add the Fixes tag.  It looks like the drop action never worked
> properly because of the vf check.

I haven't been able to verify, but based on reading the code it looks like
this path may not have worked since the code was introduced.

In that case, a fixes tag isn't needed because this is essentially new
functionality.

> > ---
> >  drivers/net/ethernet/broadcom/bnxt/bnxt_ethtool.c | 10 ++++++----
> >  1 file changed, 6 insertions(+), 4 deletions(-)
> >
> > diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt_ethtool.c b/drivers/net/ethernet/broadcom/bnxt/bnxt_ethtool.c
> > index 6b15fedbb16f..fd32231bf8e0 100644
> > --- a/drivers/net/ethernet/broadcom/bnxt/bnxt_ethtool.c
> > +++ b/drivers/net/ethernet/broadcom/bnxt/bnxt_ethtool.c
> > @@ -1353,10 +1353,12 @@ static int bnxt_add_ntuple_cls_rule(struct bnxt *bp,
> >         if (!bp->vnic_info)
> >                 return -EAGAIN;
> >
> > -       vf = ethtool_get_flow_spec_ring_vf(fs->ring_cookie);
> > -       ring = ethtool_get_flow_spec_ring(fs->ring_cookie);
> > -       if ((fs->flow_type & (FLOW_MAC_EXT | FLOW_EXT)) || vf)
> > -               return -EOPNOTSUPP;
> > +       if (fs->ring_cookie != RX_CLS_FLOW_DISC) {
> > +               vf = ethtool_get_flow_spec_ring_vf(fs->ring_cookie);
> > +               ring = ethtool_get_flow_spec_ring(fs->ring_cookie);
> > +               if ((fs->flow_type & (FLOW_MAC_EXT | FLOW_EXT)) || vf)
> 
> I think the FLOW_MAC_EXT and FLOW_EXT checks should be done unconditionally.

OK, I'll change the patch so that -EOPNOTSUPP is returned unconditionally if
(FLOW_MAC_EXT | FLOW_EXT) are set and then do the RX_CLS_FLOW_DISC flow check
before calling ethtool_get_flow_spec_ring_vf.

> Alternatively, we can also change ethtool_get_flow_spec_ring_vf() to
> check that the cookie is not RX_CLS_FLOW_DISC or RX_CLS_FLOW_WAKE.

That's up to the maintainers.

The existing drivers I checked which use ethtool_get_flow_spec_ring_vf all
explicitly check RX_CLS_FLOW_DISC, so it seems to be an established pattern.

I think we should fix this in the broadcom driver first and if there is a
desire from the maintainers to handle this in ethtool_get_flow_spec_ring_vf,
that can be done as a separate series which updates all of the drivers to
remove the redundant check.