On Thu, Aug 21, 2025 at 1:08 PM Xavier Simonart via dev <
[email protected]> wrote:
> When a packet is slow going from one hv to the other one through localnet,
> FDB
> related flows, created after the first hv updated FDB, might already be
> installed in the second hv.
> Hence, we cannot blindly expect PACKET_IN to be generated.
>
> Fixes: 2acf91e9628e ("controller: FDB entries for localnet should not
> overwrite entries for vifs")
> Fixes: 7ff2c6be5749 ("controller: Fix potentially missing vif entries in
> fdb.")
>
> Signed-off-by: Xavier Simonart <[email protected]>
> ---
> tests/ovn.at | 27 ++++++++++++++-------------
> 1 file changed, 14 insertions(+), 13 deletions(-)
>
> diff --git a/tests/ovn.at b/tests/ovn.at
> index 976ab64db..292ca0dae 100644
> --- a/tests/ovn.at
> +++ b/tests/ovn.at
> @@ -39831,6 +39831,14 @@ expected_flow_count=$(($expected_flow_count + 4))
> check_flow_count hv1 $expected_flow_count
> check_flow_count hv2 $expected_flow_count
>
> +AT_CHECK([test 5 = `cat hv1/ovs-vswitchd.log | grep NXT_PACKET_IN2 | wc
> -l`])
> +# Do not check number of PACKET_IN2 on hv2, as vif14 => 99 as well as
> vif15 => vif25
> +# might have caused PACKET_IN2, but without any guarantee.
> +# e.g. (1) packet => hv1 ovs, (2) PACKET_IN2 => hv1 ovn, (3) FDB => sb,
> +# (4) FDB => hv2 ovn, (5) flows => hv2 ovs.
> +# If packet is received by hv2 ovs before (5), there will be a PACKET_IN2.
> +# If packet is only received by hv2 ovs after (5), there will be no
> PACKET_IN2.
> +
> AS_BOX([$(date +%H:%M:%S.%03N) Packet from external])
> arp_req=$(fmt_pkt "Ether(dst='00:00:00:00:10:50',
> src='00:00:00:00:10:40')/ARP(op=2, pdst='192.168.10.50',
> psrc='192.168.10.40')")
> as hv2 ovs-appctl netdev-dummy/receive ext0 $arp_req
> @@ -39917,17 +39925,10 @@ wake_up_controller hv1
> ensure_controller_run hv1
> wait_column "$vif23_key" fdb port_key mac='"00:00:00:00:10:23"'
>
> -# In both controllers
> -# - vif11 <=> vif21: 1 PACKET_IN
> -# - vif12 <=> vif22: 2 PACKET_IN
> -# - vif13 <=> vif23: 2 PACKET_IN
> -# - vif14 <=> 99: 1 PACKET_IN
> -# - vif24 <=> vif14: 1 PACKET_IN
> -# - vif25 <=> vif15: 2 PACKET_IN
> -# - ext0: 1 PACKET_IN
> -# controller + .
> -AT_CHECK([test 10 = `cat hv1/ovs-vswitchd.log | grep NXT_PACKET_IN2 | wc
> -l`])
> -AT_CHECK([test 10 = `cat hv2/ovs-vswitchd.log | grep NXT_PACKET_IN2 | wc
> -l`])
> +# Do not check number of PACKET_IN2, as multiple packets might have
> causes PACKET_IN2,
> +# but without any guarantee.
> +n_packet_in_hv1=$(cat hv1/ovs-vswitchd.log | grep NXT_PACKET_IN2 | wc -l)
> +n_packet_in_hv2=$(cat hv2/ovs-vswitchd.log | grep NXT_PACKET_IN2 | wc -l)
>
> # Send a few more packets
> send_packet hv1 13 23
> @@ -39948,8 +39949,8 @@ for i in 1 2; do
> done
>
> # The last packets should have gone through the fast path
> -AT_CHECK([test 10 = `cat hv1/ovs-vswitchd.log | grep NXT_PACKET_IN2 | wc
> -l`])
> -AT_CHECK([test 10 = `cat hv2/ovs-vswitchd.log | grep NXT_PACKET_IN2 | wc
> -l`])
> +AT_CHECK([test $n_packet_in_hv1 = `cat hv1/ovs-vswitchd.log | grep
> NXT_PACKET_IN2 | wc -l`])
> +AT_CHECK([test $n_packet_in_hv2 = `cat hv2/ovs-vswitchd.log | grep
> NXT_PACKET_IN2 | wc -l`])
>
> # Check that there are no bad surprises
> wait_column "$vif11_key" fdb port_key mac='"00:00:00:00:10:11"'
> --
> 2.47.1
>
> _______________________________________________
> dev mailing list
> [email protected]
> https://mail.openvswitch.org/mailman/listinfo/ovs-dev
>
>
Thank you Xavier,
I went ahead and merged this into main and backported all the way down to
24.03.
Regards,
Ales
_______________________________________________
dev mailing list
[email protected]
https://mail.openvswitch.org/mailman/listinfo/ovs-dev
