| Message ID | 20230314013423.12029-1-vladimir@nikishkin.pw (mailing list archive) |
|---|---|
| State | Changes Requested |
| Delegated to: | Netdev Maintainers |
| Headers | show |
| Series | [net-next,v3] Make vxlan try to send a packet normally if local bypass fails. | expand |
On 14/03/2023 03:34, Vladimir Nikishkin wrote: > In vxlan_core, if an fdb entry is pointing to a local > address with some port, the system tries to get the packet to > deliver the packet to the vxlan directly, bypassing the network > stack. > > This patch makes it still try canonical delivery, if there is no > linux kernel vxlan listening on this port. This will be useful > for the cases when there is some userspace daemon expecting > vxlan packets for post-processing, or some other implementation > of vxlan. > > Signed-off-by: Vladimir Nikishkin <vladimir@nikishkin.pw> > --- > drivers/net/vxlan/vxlan_core.c | 10 ++-------- > 1 file changed, 2 insertions(+), 8 deletions(-) > Hi Vladimir, You should structure the subject to driver: change, in this case you can use something like vxlan: try to send a packet normally if local bypass fails > diff --git a/drivers/net/vxlan/vxlan_core.c b/drivers/net/vxlan/vxlan_core.c > index b1b179effe2a..0379902da766 100644 > --- a/drivers/net/vxlan/vxlan_core.c > +++ b/drivers/net/vxlan/vxlan_core.c > @@ -2422,19 +2422,13 @@ static int encap_bypass_if_local(struct sk_buff *skb, struct net_device *dev, > if (rt_flags & RTCF_LOCAL && > !(rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) { > struct vxlan_dev *dst_vxlan; > - > - dst_release(dst); > dst_vxlan = vxlan_find_vni(vxlan->net, dst_ifindex, vni, > daddr->sa.sa_family, dst_port, > vxlan->cfg.flags); > if (!dst_vxlan) { you can drop the {} for a single line > - dev->stats.tx_errors++; > - vxlan_vnifilter_count(vxlan, vni, NULL, > - VXLAN_VNI_STATS_TX_ERRORS, 0); > - kfree_skb(skb); > - > - return -ENOENT; > + return 0; > } > + dst_release(dst); > vxlan_encap_bypass(skb, vxlan, dst_vxlan, vni, true); > return 1; > } This also changes such packet delivery expectations so these packets are not dropped anymore which might surprise people and allow traffic that would've been dropped otherwise. IMO it should be hidden behind a new vxlan option that defaults to the old behaviour. Cheers, Nik
diff --git a/drivers/net/vxlan/vxlan_core.c b/drivers/net/vxlan/vxlan_core.c index b1b179effe2a..0379902da766 100644 --- a/drivers/net/vxlan/vxlan_core.c +++ b/drivers/net/vxlan/vxlan_core.c @@ -2422,19 +2422,13 @@ static int encap_bypass_if_local(struct sk_buff *skb, struct net_device *dev, if (rt_flags & RTCF_LOCAL && !(rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) { struct vxlan_dev *dst_vxlan; - - dst_release(dst); dst_vxlan = vxlan_find_vni(vxlan->net, dst_ifindex, vni, daddr->sa.sa_family, dst_port, vxlan->cfg.flags); if (!dst_vxlan) { - dev->stats.tx_errors++; - vxlan_vnifilter_count(vxlan, vni, NULL, - VXLAN_VNI_STATS_TX_ERRORS, 0); - kfree_skb(skb); - - return -ENOENT; + return 0; } + dst_release(dst); vxlan_encap_bypass(skb, vxlan, dst_vxlan, vni, true); return 1; }
In vxlan_core, if an fdb entry is pointing to a local address with some port, the system tries to get the packet to deliver the packet to the vxlan directly, bypassing the network stack. This patch makes it still try canonical delivery, if there is no linux kernel vxlan listening on this port. This will be useful for the cases when there is some userspace daemon expecting vxlan packets for post-processing, or some other implementation of vxlan. Signed-off-by: Vladimir Nikishkin <vladimir@nikishkin.pw> --- drivers/net/vxlan/vxlan_core.c | 10 ++-------- 1 file changed, 2 insertions(+), 8 deletions(-)