RE: [Intel-wired-lan] [PATCH iwl-next v4 6/6] ice: enable LLDP TX for VFs through tc

[Romanowski, Rafal]

> -----Original Message-----
> From: Intel-wired-lan <intel-wired-lan-bounces@xxxxxxxxxx> On Behalf Of
> Simon Horman
> Sent: Thursday, February 20, 2025 3:59 PM
> To: Zaremba, Larysa <larysa.zaremba@xxxxxxxxx>
> Cc: Nguyen, Anthony L <anthony.l.nguyen@xxxxxxxxx>; intel-wired-
> lan@xxxxxxxxxxxxxxxx; Kitszel, Przemyslaw <przemyslaw.kitszel@xxxxxxxxx>;
> Andrew Lunn <andrew+netdev@xxxxxxx>; David S. Miller
> <davem@xxxxxxxxxxxxx>; Eric Dumazet <edumazet@xxxxxxxxxx>; Jakub
> Kicinski <kuba@xxxxxxxxxx>; Paolo Abeni <pabeni@xxxxxxxxxx>;
> netdev@xxxxxxxxxxxxxxx; linux-kernel@xxxxxxxxxxxxxxx; Michal Swiatkowski
> <michal.swiatkowski@xxxxxxxxxxxxxxx>; Pacuszka, MateuszX
> <mateuszx.pacuszka@xxxxxxxxx>
> Subject: Re: [Intel-wired-lan] [PATCH iwl-next v4 6/6] ice: enable LLDP TX for
> VFs through tc
> 
> On Fri, Feb 14, 2025 at 09:50:40AM +0100, Larysa Zaremba wrote:
> > Only a single VSI can be in charge of sending LLDP frames, sometimes
> > it is beneficial to assign this function to a VF, that is possible to
> > do with tc capabilities in the switchdev mode. It requires first
> > blocking the PF from sending the LLDP frames with a following command:
> >
> > tc filter add dev <ifname> egress protocol lldp flower skip_sw action
> > drop
> >
> > Then it becomes possible to configure a forward rule from a VF port
> > representor to uplink instead.
> >
> > tc filter add dev <vf_ifname> ingress protocol lldp flower skip_sw
> > action mirred egress redirect dev <ifname>
> >
> > How LLDP exclusivity was done previously is LLDP traffic was blocked
> > for a whole port by a single rule and PF was bypassing that. Now at
> > least in the switchdev mode, every separate VSI has to have its own
> > drop rule. Another complication is the fact that tc does not respect
> > when the driver refuses to delete a rule, so returning an error
> > results in a HW rule still present with no way to reference it through
> > tc. This is addressed by allowing the PF rule to be deleted at any time, but
> making the VF forward rule "dormant"
> > in such case, this means it is deleted from HW but stays in tc and
> > driver's bookkeeping to be restored when drop rule is added back to the PF.
> >
> > Implement tc configuration handling which enables the user to transmit
> > LLDP packets from VF instead of PF.
> >
> > Reviewed-by: Michal Swiatkowski <michal.swiatkowski@xxxxxxxxxxxxxxx>
> > Signed-off-by: Larysa Zaremba <larysa.zaremba@xxxxxxxxx>
> 
> Reviewed-by: Simon Horman <horms@xxxxxxxxxx>

Tested-by: Rafal Romanowski <rafal.romanowski@xxxxxxxxx>