luci-mod-network: link bonding policy not propagated

[jpsollie]

Is there an existing issue for this?

• I have searched among all existing issues (including closed issues)

screenshots or captures

when creating a link aggregation interface with the new netifd system instead of luci-proto-bonding (which was removed a few weeks ago), I enter in luci the option "active-backup" (view screenshot):

But in /proc/net/bondjp it states round-robin:

cat /proc/net/bonding/bondjp 
Ethernet Channel Bonding Driver: v6.6.83

Bonding Mode: load balancing (round-robin)
MII Status: up
MII Polling Interval (ms): 0
Up Delay (ms): 0
Down Delay (ms): 0
Peer Notification Delay (ms): 0

Slave Interface: lan1
MII Status: up
Speed: 1000 Mbps
Duplex: full
Link Failure Count: 0
Permanent HW addr: ba:96:7a:2e:1b:71
Slave queue ID: 0

Slave Interface: sfp-lan
MII Status: up
Speed: 2500 Mbps
Duplex: full
Link Failure Count: 0
Permanent HW addr: ba:96:7a:2e:1b:72
Slave queue ID: 0
root@APBureau4:~# 

There's also nothing written in /etc/config/networks about an active-backup policy (though I don't know whether this should be, the documentation isn't clear):

config device
        option type 'bonding'
        option name 'bondjp'
        list ports 'lan1'
        list ports 'sfp-lan'
        option primary 'sfp-lan'
        option multicast_router '2'
        option multicast_fast_leave '1'
        option promisc '1'

IMO this is a bug, and I hope the missing parameter can be added, but if someone could help me to verify whether uci knows it should be active-backup (which makes it a main openwrt bug), it'd love to hear

Actual behaviour

a newly created link aggegation interface is started in round-robin mode, whereas it should be as active-backup

Expected behaviour

the interface is started as active-backup

Steps to reproduce

Additional Information

{
        "kernel": "6.6.83",
        "hostname": "APBureau4",
        "system": "ARMv8 Processor rev 0",
        "model": "Bananapi BPI-R4",
        "board_name": "bananapi,bpi-r4",
        "rootfs_type": "squashfs",
        "release": {
                "distribution": "OpenWrt",
                "version": "SNAPSHOT",
                "firmware_url": "https://downloads.openwrt.org/",
                "revision": "r0+29008-0aaabffdea",
                "target": "mediatek/filogic",
                "description": "OpenWrt SNAPSHOT r0+29008-0aaabffdea",
                "builddate": "1742051600"
        }
}

What browsers do you see the problem on?

No response

Relevant log output

[systemcrash]

Since active-backup is the default, perhaps it is elided, with the assumption that netifd also chooses that.

Please try to add the explicit config line to /etc/config/networks

policy 'active-backup'

resulting in:

config device
        option type 'bonding'
        option name 'bondjp'
        list ports 'lan1'
        list ports 'sfp-lan'
        option primary 'sfp-lan'
        option multicast_router '2'
        option multicast_fast_leave '1'
        option promisc '1'
        policy 'active-backup'

and reload that interface or restart networking. Let us know your results.

[systemcrash]

Any ideas from @hitech95 ?

[jpsollie]

policy active-backup didn't do what we expected it would do ... do I need to forward this bug to the main repository?

[hitech95]

Any ideas from @hitech95 ?

I'll take a look later today.
Thi might be a bug in both luci and netifd.

policy active-backup didn't do what we expected it would do ... do I need to forward this bug to the main repository?

Can you better explain the expected result so I can later run a test on your config?
What was the configuration with luci-proto-bonding?

[jpsollie]

the expected result would be:
cat /proc/net/bond/bondjp shows "Bonding Mode: Active-backup"
Attached I'll add my /etc/config/networks with luci-proto-bonding

network.txt

[systemcrash]

At a glance it seems the variable simply needs renaming to bonding_policy. @jpsollie please try:

config device
        option type 'bonding'
        option name 'bondjp'
        list ports 'lan1'
        list ports 'sfp-lan'
        option primary 'sfp-lan'
        option multicast_router '2'
        option multicast_fast_leave '1'
        option promisc '1'
        bonding_policy 'active-backup'

[jpsollie]

At a glance it seems the variable simply needs renaming to bonding_policy. @jpsollie please try:

config device
        option type 'bonding'
        option name 'bondjp'
        list ports 'lan1'
        list ports 'sfp-lan'
        option primary 'sfp-lan'
        option multicast_router '2'
        option multicast_fast_leave '1'
        option promisc '1'
        bonding_policy 'active-backup'

nope, didn't work, still round-robin

[systemcrash]

OK.

https://github.com/openwrt/netifd/blob/ea01ed41f3212ecbe000422f3c122a01b93fe874/bonding.c#L86-L112

Source says policy.

Might it be worth testing to see whether the machinery is in place and manually setting bonding itself works with various modes?:

ip link add bond0 type bond
ip link set bond0 down
ip link set bond0 type bond mode <mode>
ip link set lan1 down
ip link set lan1 master bond0
ip link set sfp-lan down
ip link set sfp-lan master bond0
ip link set bond0 up

[hitech95]

I confirm tha the node is called 'policy' yesterday I've missed that.

[jpsollie]

OK.

https://github.com/openwrt/netifd/blob/ea01ed41f3212ecbe000422f3c122a01b93fe874/bonding.c#L86-L112

Source says policy.

Might it be worth testing to see whether the machinery is in place and manually setting bonding itself works with various modes?:

ip link add bond0 type bond
ip link set bond0 down
ip link set bond0 type bond mode <mode>
ip link set lan1 down
ip link set lan1 master bond0
ip link set sfp-lan down
ip link set sfp-lan master bond0
ip link set bond0 up

ok, I put your commands in a shell script:

#!/bin/bash
ip link add bond0 type bond
ip link set bond0 down
ip link set bond0 type bond mode active-backup
ip link set lan1 down
ip link set lan1 master bond0
ip link set sfp-lan down
ip link set sfp-lan master bond0
ip link set bond0 up

and executed it, here's what I got:

root@APBureau4:~# ip link del bondjp
root@APBureau4:~# sh /tmp/shellscript.sh
root@APBureau4:~# cat /proc/net/b
bnep      bonding/
root@APBureau4:~# cat /proc/net/bonding/bond0
Ethernet Channel Bonding Driver: v6.6.83

Bonding Mode: fault-tolerance (active-backup)
Primary Slave: None
Currently Active Slave: lan1
MII Status: up
MII Polling Interval (ms): 0
Up Delay (ms): 0
Down Delay (ms): 0
Peer Notification Delay (ms): 0

Slave Interface: lan1
MII Status: up
Speed: Unknown
Duplex: Unknown
Link Failure Count: 0
Permanent HW addr: ba:96:7a:2e:1b:71
Slave queue ID: 0

Slave Interface: sfp-lan
MII Status: up
Speed: 2500 Mbps
Duplex: full
Link Failure Count: 0
Permanent HW addr: ba:96:7a:2e:1b:72
Slave queue ID: 0
root@APBureau4:~#

so I guess that's what it should be

[systemcrash]

It looks like the system can do it correctly. Not necessarily netifd. But when you used the proto handler @jpsollie , link bonding also worked?

[jpsollie]

I assume it did (don't take my word for it): I never had problems with active-backup before, and the policy was set to active-backup on the client PC's networkmanager ...