selftests: bonding: disable DAD for IPv6 addresses
Subsystems (4)
Jakub Kicinski
maintainerPATCH 1/1 selftests: bonding: disable DAD for IPv6 addresses
bond_reset() waits up to 2 seconds for IPv6 connectivity. With default settings DAD itself may take almost 2 seconds, causing flakes on debug builds. It used to flake once or twice a week, recently it started failing once a day. Probably some downstream changes to scheduler, or our machines go busier.
A lot of selftests already use nodad, let's use nodad in bonding, too. I don't see an obvious reason why DAD would be important to the test.
---
CC: [email protected]
CC: [email protected]
CC: [email protected]
CC: [email protected]
--- .../selftests/drivers/net/bonding/bond_topo_2d1c.sh | 10 ++++------ 1 file changed, 4 insertions(+), 6 deletions(-)
tools/testing/selftests/drivers/net/bonding/bond_topo_2d1c.sh +4 -6
@@ -48,7 +48,7 @@ gateway_create() ip -n ${g_ns} link add br0 type bridge ip -n ${g_ns} link set br0 up ip -n ${g_ns} addr add ${g_ip4}/24 dev br0 - ip -n ${g_ns} addr add ${g_ip6}/24 dev br0 + ip -n ${g_ns} addr add ${g_ip6}/24 dev br0 nodad } gateway_destroy()
@@ -75,7 +75,7 @@ server_create() ip -n ${s_ns} link set bond0 up ip -n ${s_ns} addr add ${s_ip4}/24 dev bond0 - ip -n ${s_ns} addr add ${s_ip6}/24 dev bond0 + ip -n ${s_ns} addr add ${s_ip6}/24 dev bond0 nodad } # Reset bond with new mode and options
@@ -97,9 +97,7 @@ bond_reset() ip -n ${s_ns} link set bond0 up ip -n ${s_ns} addr add ${s_ip4}/24 dev bond0 - ip -n ${s_ns} addr add ${s_ip6}/24 dev bond0 - # Wait for IPv6 address ready as it needs DAD - slowwait 2 ip netns exec ${s_ns} ping6 ${c_ip6} -c 1 -W 0.1 &> /dev/null + ip -n ${s_ns} addr add ${s_ip6}/24 dev bond0 nodad } server_destroy()
@@ -124,7 +122,7 @@ client_create() ip -n ${c_ns} link set eth0 up ip -n ${c_ns} addr add ${c_ip4}/24 dev eth0 - ip -n ${c_ns} addr add ${c_ip6}/24 dev eth0 + ip -n ${c_ns} addr add ${c_ip6}/24 dev eth0 nodad } client_destroy()
Signature · 1 line
2.55.0
Quoted 9 lines · Jakub Kicinski
bond_reset()waits up to 2 seconds for IPv6 connectivity. With default settings DAD itself may take almost 2 seconds, causing flakes on debug builds. It used to flake once or twice a week, recently it started failing once a day. Probably some downstream changes to scheduler, or our machines go busier.A lot of selftests already use nodad, let's use nodad in bonding, too. I don't see an obvious reason why DAD would be important to the test.
In this case, how about add nodad to all IPv6 address setting on bonding tests, e.g. bond_options.sh, bond_macvlan_ipvlan.sh.
After that, we may event remove the sleep time in bond_macvlan_ipvlan.sh. In my local testing all the macvlan/ipvlan test passed with nodad set.
Thanks
Hangbin
Quoted 56 lines
Signed-off-by Jakub Kicinski <[email protected]>---
CC: [email protected]
CC: [email protected]
CC: [email protected]
CC: [email protected]--- .../selftests/drivers/net/bonding/bond_topo_2d1c.sh | 10 ++++------ 1 file changed, 4 insertions(+), 6 deletions(-)tools/testing/selftests/drivers/net/bonding/bond_topo_2d1c.sh +4
-6@@ -48,7 +48,7 @@ gateway_create() ip -n ${g_ns} link add br0 type bridge ip -n ${g_ns} link set br0 up ip -n ${g_ns} addr add ${g_ip4}/24 dev br0 - ip -n ${g_ns} addr add ${g_ip6}/24 dev br0 + ip -n ${g_ns} addr add ${g_ip6}/24 dev br0 nodad } gateway_destroy()@@ -75,7 +75,7 @@ server_create() ip -n ${s_ns} link set bond0 up ip -n ${s_ns} addr add ${s_ip4}/24 dev bond0 - ip -n ${s_ns} addr add ${s_ip6}/24 dev bond0 + ip -n ${s_ns} addr add ${s_ip6}/24 dev bond0 nodad } # Reset bond with new mode and options@@ -97,9 +97,7 @@ bond_reset() ip -n ${s_ns} link set bond0 up ip -n ${s_ns} addr add ${s_ip4}/24 dev bond0 - ip -n ${s_ns} addr add ${s_ip6}/24 dev bond0 - # Wait for IPv6 address ready as it needs DAD - slowwait 2 ip netns exec ${s_ns} ping6 ${c_ip6} -c 1 -W 0.1 &> /dev/null + ip -n ${s_ns} addr add ${s_ip6}/24 dev bond0 nodad } server_destroy()@@ -124,7 +122,7 @@ client_create() ip -n ${c_ns} link set eth0 up ip -n ${c_ns} addr add ${c_ip4}/24 dev eth0 - ip -n ${c_ns} addr add ${c_ip6}/24 dev eth0 + ip -n ${c_ns} addr add ${c_ip6}/24 dev eth0 nodad } client_destroy()Signature · 1 line
2.55.0
Jakub Kicinski
maintainerQuoted 8 lines · On Mon, 10 Aug 2026 10:02:10 +0800 Hangbin Liu wrote:
Quoted 7 lines · Jakub Kicinski
bond_reset()waits up to 2 seconds for IPv6 connectivity. With default settings DAD itself may take almost 2 seconds, causing flakes on debug builds. It used to flake once or twice a week, recently it started failing once a day. Probably some downstream changes to scheduler, or our machines go busier.
Quoted 8 lines
Quoted 2 lines
A lot of selftests already use nodad, let's use nodad in bonding, too. I don't see an obvious reason why DAD would be important to the test.
In this case, how about add nodad to all IPv6 address setting on bonding tests, e.g.
bond_options.sh,bond_macvlan_ipvlan.sh.After that, we may event remove the sleep time in
bond_macvlan_ipvlan.sh. In my local testing all the macvlan/ipvlan test passed with nodad set.
Can you do that as a follow up, please?
bond_options.sh is the worst offender for flakiness right now. I would like to squash these flakes ASAP.
Hangbin Liu
Reviewed-byQuoted 10 lines · Jakub Kicinski
Quoted 5 lines
In this case, how about add nodad to all IPv6 address setting on bonding tests, e.g.
bond_options.sh,bond_macvlan_ipvlan.sh.After that, we may event remove the sleep time in
bond_macvlan_ipvlan.sh. In my local testing all the macvlan/ipvlan test passed with nodad set.Can you do that as a follow up, please?
bond_options.shis the worst offender for flakiness right now. I would like to squash these flakes ASAP.
Sure, and in this case
This patch was applied to netdev/net-next.git (main) by Jakub Kicinski <[email protected]>:
Hello:
This patch was applied to netdev/net-next.git (main) by Jakub Kicinski <[email protected]>:
Quoted 8 lines · On Sat, 8 Aug 2026 09:23:45 -0700 you wrote:
bond_reset()waits up to 2 seconds for IPv6 connectivity. With default settings DAD itself may take almost 2 seconds, causing flakes on debug builds. It used to flake once or twice a week, recently it started failing once a day. Probably some downstream changes to scheduler, or our machines go busier.[...]
Here is the summary with links:
- [net-next] selftests: bonding: disable DAD for IPv6 addresses
https://git.kernel.org/netdev/net-next/c/30dbc21f6375
You are awesome, thank you!
Signature · 2 lines
Deet-doot-dot, I am a bot. https://korg.docs.kernel.org/patchwork/pwbot.html