Batman stopped working on Ath79 snapshot after last batch of mac80211 updates

stragies · August 31, 2021, 1:07am

@nbd Up to current git master head (r17467 , 5.4.143) I have isolated the following block of commits, that break batman since mid-august:

0f6887972a mac80211: add missing change for encap offload on devices with sw rate control
f04c0ead33 mac80211: refresh patch
a0d81ba0d5 mac80211: fix HT40 mode for 6G band

The symptoms are that batctl n shows neighbors, but batctl o shows no originators, batctl ping fails to/from all shown neighbors, and no mesh traffic goes through bat0. Also the other devices do not even see this device as a neigbour. The actual physical interface works fine, I was able to to SSH to the IPv6 allocated to the Hard_IF below batman.

I noticed, when my on-desk test-device (CONFIG_TARGET_ath79_generic_DEVICE_glinet_gl-ar300m-lite=y) started exhibiting this problem. Other devices still working had 5.4.142, so the problem had to be after that. git reverting the above 3 commits gets Batman working again normally.

markbirss · August 31, 2021, 9:47am

Can you ping the ip address of the other node ?

It seems the bridge definition part has changed in latest version of OpenWRT
How have you defined your lan mesh bridge in /etc/config/network

stragies · August 31, 2021, 10:26am

Can you ping the ip address of the other node ?

Mesh-nodes can ping6 the IPv6 of each-others mesh2g-interface, yes.
None of the other meshnodes can batctl p the (running snapshot) AR300m. In fact, they don't even see the AR300 in the batctl n neighbors list.

All magically goes back to normal when using the 21.02.0-rc4.

It seems the bridge definition part has changed in latest version of OpenWRT

I don't understand, how this comes into play at this level. Or are you not talking about the "LAN"-bridge? In my case LAN is a bridge with eth0 and bat0.myVLANNumber. IP-traffic to/from other devices on LAN-L2 does not work.

markbirss · August 31, 2021, 10:38am

Ok, i had strange behaviour on a MT7620 as soon as i enabled encryption, only worked after i disabled encryption

Does the throughput stay at 1 Mbps when you run iw station dump perhaps ?

stragies · August 31, 2021, 10:44am

I have MT76 devices as well. Will also test them.

No, station dump looks "normal", i think:

Station 12:34:56:78:90:de (on mesh2g)
        inactive time:  10 ms
        rx bytes:       1246718
        rx packets:     7032
        tx bytes:       118026
        tx packets:     547
        tx retries:     161
        tx failed:      0
        rx drop misc:   54
        signal:         -40 [-45, -42] dBm
        signal avg:     -41 [-46, -43] dBm
        Toffset:        775307160489 us
        tx bitrate:     130.0 MBit/s MCS 15
        tx duration:    52663 us
        rx bitrate:     144.4 MBit/s MCS 15 short GI
        rx duration:    10664262 us
        airtime weight: 256
        expected throughput:    45.226Mbps
        mesh llid:      0
        mesh plid:      0
        mesh plink:     ESTAB
        mesh airtime link metric: 177
        mesh connected to gate: no
        mesh connected to auth server:  no
        mesh local PS mode:     ACTIVE
        mesh peer PS mode:      ACTIVE
        mesh non-peer PS mode:  ACTIVE
        authorized:     yes
        authenticated:  yes
        associated:     yes
        preamble:       long
        WMM/WME:        yes
        MFP:            yes
        TDLS peer:      no
        DTIM period:    2
        beacon interval:5000
        connected time: 475 seconds
        associated at [boottime]:       37.992s
        associated at:  1630389169897 ms
        current time:   1630389644741 ms

markbirss · August 31, 2021, 10:47am

Ok, dont know what else to suggest.

I see you are installing alfred

Do you use luci-app-batman-adv ?

stragies · August 31, 2021, 10:54am

Thanks in any case

Uhm, yes, for testing. But not in the images built with imagebuilder. So it makes no difference in this case.

Where would I find that? I don't have that option/app in my menuconfig. I searched through / for "luci-app-b" and "batman".

markbirss · August 31, 2021, 11:18am

Since i was also building LibreMesh images, I took this ipk from those builds

LibreMesh only builds successfully upto OpenWRT 19.07.7

it a universal ipk not target specific

luci-app-batman-adv_2021-07-08-1625782035_all.ipk

You then have a luci front-end for visualisation of your batman nodes

I was also running batman-adv on ethernet also

markbirss · August 31, 2021, 11:55am

I built the package

markbirss · August 31, 2021, 12:57pm

CONFIG_PACKAGE_luci-app-batman-adv=y

stragies · August 31, 2021, 4:58pm

In slightly more detail:

git clone https://github.com/libremesh/lime-packages somewhere
cp the subfolder luci-app-batman-adv into the openwrt as feeds/packages/net/luci-app-batman-adv/
update & install the package feeds
make menuconfig
Luci -> Applications -> luci-app-batadv

So, now I have the Luci Batman GUI, that tells me the same thing: Neigbors can be seen, but originators stays at zero. My devices still running 5.4.142 (r17390+11-9baca41064) from 10 days ago still have functioning batman. It broke somewhere since then.

Update: The three last mac80211 patches (between 5.4.142 and .143) seem to provoke this problem here.

nbd · September 1, 2021, 1:17pm

@stragies can you please test if reverting only 0f6887972a and not the other two will also make it work again?

stragies · September 1, 2021, 2:08pm

I thought, I had already tried that last night ... that's why I wrote, that reverting all 3 is required ... Oops.

Yes, reverting only 0f6887972a suffices to get Batman working again properly. Sorry for the confusion.

stragies · September 14, 2021, 7:52am

Hi @nbd, just to confirm: My builds since the revised patch you committed upstream do not exhibit the problem anymore. Thanks!

system · September 24, 2021, 7:52am

This topic was automatically closed 10 days after the last reply. New replies are no longer allowed.