Message ID | 1575989382-7195-1-git-send-email-xiangxia.m.yue@gmail.com |
---|---|
State | Awaiting Upstream |
Delegated to: | David Miller |
Headers | show |
Series | [net-next] net/mlx5e: Support accept action on nic table | expand |
On 12/10/19 6:49 AM, xiangxia.m.yue@gmail.com wrote: > From: Tonghao Zhang <xiangxia.m.yue@gmail.com> > > In one case, we may forward packets from one vport > to others, but only one packets flow will be accepted, > which destination ip was assign to VF. > > +-----+ +-----+ +-----+ > | VFn | | VF1 | | VF0 | accept > +--+--+ +--+--+ hairpin +--^--+ > | | <--------------- | > | | | > +--+-----------v-+ +--+-------------+ > | eswitch PF1 | | eswitch PF0 | > +----------------+ +----------------+ > > tc filter add dev $PF0 protocol all parent ffff: prio 1 handle 1 \ > flower skip_sw action mirred egress redirect dev $VF0_REP > tc filter add dev $VF0 protocol ip parent ffff: prio 1 handle 1 \ > flower skip_sw dst_ip $VF0_IP action pass > tc filter add dev $VF0 protocol all parent ffff: prio 2 handle 2 \ > flower skip_sw action mirred egress redirect dev $VF1 > > Signed-off-by: Tonghao Zhang <xiangxia.m.yue@gmail.com> > --- > drivers/net/ethernet/mellanox/mlx5/core/en_tc.c | 4 ++++ > 1 file changed, 4 insertions(+) > > diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_tc.c b/drivers/net/ethernet/mellanox/mlx5/core/en_tc.c > index 0d5d84b..f91e057e 100644 > --- a/drivers/net/ethernet/mellanox/mlx5/core/en_tc.c > +++ b/drivers/net/ethernet/mellanox/mlx5/core/en_tc.c > @@ -2839,6 +2839,10 @@ static int parse_tc_nic_actions(struct mlx5e_priv *priv, > > flow_action_for_each(i, act, flow_action) { > switch (act->id) { > + case FLOW_ACTION_ACCEPT: > + action |= MLX5_FLOW_CONTEXT_ACTION_FWD_DEST | > + MLX5_FLOW_CONTEXT_ACTION_COUNT; > + break; > case FLOW_ACTION_DROP: > action |= MLX5_FLOW_CONTEXT_ACTION_DROP; > if (MLX5_CAP_FLOWTABLE(priv->mdev, > Roi once Saeed takes it, can you please make sure we are adding a test for this flow? a (very) simple (basic) way to verify this flow is to do something like this: 1) move both PFs into switchdev VF0_PF0=ens4f2 VF0_PF0_REP=ens4f0_0 VF0_PF1=ens4f5 VF0_PF1_REP=ens4f1_0 ethtool -K VF0_PF0 hw-tc-offload on tc filter add dev VF0_PF0 protocol ip parent ffff: prio 1 handle 1 flower skip_sw dst_ip 1.1.1.1 action pass tc filter add dev VF0_PF0 protocol ip parent ffff: prio 2 handle 2 flower skip_sw action mirred egress redirect dev VF0_PF1 ifconfig VF0_PF0_REP 1.1.1.2 arp -s 1.1.1.1 ec:0d:9a:d4:2e:14 arp -s 1.1.1.3 ec:0d:9a:d4:2e:24 Now if you try to ping 1.1.1.1 you should see packets on VF0_PF0 if you try to ping 1.1.1.3 you should see packets on VF0_PF1_REP Reviewed-by: Mark Bloch <markb@mellanox.com> Mark
On Tue, 2019-12-10 at 22:49 +0800, xiangxia.m.yue@gmail.com wrote: > From: Tonghao Zhang <xiangxia.m.yue@gmail.com> > > In one case, we may forward packets from one vport > to others, but only one packets flow will be accepted, > which destination ip was assign to VF. > > +-----+ +-----+ +-----+ > > VFn | | VF1 | | VF0 | accept > +--+--+ +--+--+ hairpin +--^--+ > | | <--------------- | > | | | > +--+-----------v-+ +--+-------------+ > > eswitch PF1 | | eswitch PF0 | > +----------------+ +----------------+ > > tc filter add dev $PF0 protocol all parent ffff: prio 1 handle 1 \ > flower skip_sw action mirred egress redirect dev $VF0_REP > tc filter add dev $VF0 protocol ip parent ffff: prio 1 handle 1 \ > flower skip_sw dst_ip $VF0_IP action pass > tc filter add dev $VF0 protocol all parent ffff: prio 2 handle 2 \ > flower skip_sw action mirred egress redirect dev $VF1 > > Signed-off-by: Tonghao Zhang <xiangxia.m.yue@gmail.com> > --- > drivers/net/ethernet/mellanox/mlx5/core/en_tc.c | 4 ++++ > 1 file changed, 4 insertions(+) > > diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_tc.c > b/drivers/net/ethernet/mellanox/mlx5/core/en_tc.c > index 0d5d84b..f91e057e 100644 > --- a/drivers/net/ethernet/mellanox/mlx5/core/en_tc.c > +++ b/drivers/net/ethernet/mellanox/mlx5/core/en_tc.c > @@ -2839,6 +2839,10 @@ static int parse_tc_nic_actions(struct > mlx5e_priv *priv, > > flow_action_for_each(i, act, flow_action) { > switch (act->id) { > + case FLOW_ACTION_ACCEPT: > + action |= MLX5_FLOW_CONTEXT_ACTION_FWD_DEST | > + MLX5_FLOW_CONTEXT_ACTION_COUNT; > + break; > case FLOW_ACTION_DROP: > action |= MLX5_FLOW_CONTEXT_ACTION_DROP; > if (MLX5_CAP_FLOWTABLE(priv->mdev, Applied to net-next-mlx5, will submit to net-next soon. thanks a lot
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_tc.c b/drivers/net/ethernet/mellanox/mlx5/core/en_tc.c index 0d5d84b..f91e057e 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/en_tc.c +++ b/drivers/net/ethernet/mellanox/mlx5/core/en_tc.c @@ -2839,6 +2839,10 @@ static int parse_tc_nic_actions(struct mlx5e_priv *priv, flow_action_for_each(i, act, flow_action) { switch (act->id) { + case FLOW_ACTION_ACCEPT: + action |= MLX5_FLOW_CONTEXT_ACTION_FWD_DEST | + MLX5_FLOW_CONTEXT_ACTION_COUNT; + break; case FLOW_ACTION_DROP: action |= MLX5_FLOW_CONTEXT_ACTION_DROP; if (MLX5_CAP_FLOWTABLE(priv->mdev,