wifi_roaming makes APs unresponsive

[noxilixon]

wifi_roaming seems to cause problems at some access points, resulting in out of memory errors and unusable devices.

Location: newyorck
Device: siemens_ws-ap3610

When wifi_roaming was enabled it worked at first, but after some time I got the following errors:

At some APs I was not able to read the logs:

root@newyorck-ap-1b:~# logread 
Failed to connect to ubus

For some I was not able to make a ssh connection and needed to do a power cycle. Or I was not able to make the sysupgrade. (Again power cycle was the solution.)

root@newyorck-ap-1f:~# sysupgrade /tmp/newyorck-ap-1f.bin 
/sbin/sysupgrade: line 313: can't fork: Out of memory

Also some logs were full of this:

Wed Apr 24 12:27:55 2024 daemon.notice hostapd: nl80211: send_and_recv->nl_recvmsgs failed: -5 (Out of memory)
Wed Apr 24 12:27:55 2024 daemon.notice hostapd: nl80211: send_and_recv->nl_recvmsgs failed: -5 (Out of memory)
Wed Apr 24 12:27:55 2024 daemon.notice hostapd: nl80211: send_and_recv->nl_recvmsgs failed: -5 (Out of memory)
Wed Apr 24 12:27:55 2024 daemon.notice hostapd: nl80211: send_and_recv->nl_recvmsgs failed: -5 (Out of memory)
Wed Apr 24 12:27:55 2024 daemon.notice hostapd: nl80211: send_and_recv->nl_recvmsgs failed: -5 (Out of memory)
Wed Apr 24 12:27:55 2024 daemon.notice hostapd: nl80211: send_and_recv->nl_recvmsgs failed: -5 (Out of memory)
Wed Apr 24 12:27:55 2024 daemon.notice hostapd: nl80211: send_and_recv->nl_recvmsgs failed: -5 (Out of memory)
Wed Apr 24 12:27:55 2024 daemon.notice hostapd: nl80211: send_and_recv->nl_recvmsgs failed: -5 (Out of memory)
Wed Apr 24 12:27:55 2024 daemon.notice hostapd: nl80211: send_and_recv->nl_recvmsgs failed: -5 (Out of memory)
Wed Apr 24 12:27:55 2024 daemon.notice hostapd: nl80211: send_and_recv->nl_recvmsgs failed: -5 (Out of memory)
Wed Apr 24 12:27:55 2024 daemon.notice hostapd: nl80211: send_and_recv->nl_recvmsgs failed: -5 (Out of memory)
Wed Apr 24 12:27:55 2024 daemon.notice hostapd: nl80211: send_and_recv->nl_recvmsgs failed: -5 (Out of memory)
Wed Apr 24 12:27:55 2024 daemon.notice hostapd: nl80211: send_and_recv->nl_recvmsgs failed: -5 (Out of memory)

Or others full of this:

Wed Apr 24 16:49:56 2024 daemon.err dawn: ubus_call_umdns()=ubus.c@1385 Failed to look up test object for umdns
Wed Apr 24 16:50:12 2024 daemon.err dawn: ubus_call_umdns()=ubus.c@1385 Failed to look up test object for umdns
Wed Apr 24 16:50:28 2024 daemon.err dawn: ubus_call_umdns()=ubus.c@1385 Failed to look up test object for umdns
Wed Apr 24 16:50:44 2024 daemon.err dawn: ubus_call_umdns()=ubus.c@1385 Failed to look up test object for umdns
Wed Apr 24 16:51:00 2024 daemon.err dawn: ubus_call_umdns()=ubus.c@1385 Failed to look up test object for umdns
Wed Apr 24 16:51:16 2024 daemon.err dawn: ubus_call_umdns()=ubus.c@1385 Failed to look up test object for umdns
Wed Apr 24 16:51:32 2024 daemon.err dawn: ubus_call_umdns()=ubus.c@1385 Failed to look up test object for umdns
Wed Apr 24 16:51:48 2024 daemon.err dawn: ubus_call_umdns()=ubus.c@1385 Failed to look up test object for umdns
Wed Apr 24 16:52:04 2024 daemon.err dawn: ubus_call_umdns()=ubus.c@1385 Failed to look up test object for umdns
Wed Apr 24 16:52:20 2024 daemon.err dawn: ubus_call_umdns()=ubus.c@1385 Failed to look up test object for umdns
Wed Apr 24 16:52:36 2024 daemon.err dawn: ubus_call_umdns()=ubus.c@1385 Failed to look up test object for umdns
Wed Apr 24 16:52:52 2024 daemon.err dawn: ubus_call_umdns()=ubus.c@1385 Failed to look up test object for umdns
Wed Apr 24 16:53:08 2024 daemon.err dawn: ubus_call_umdns()=ubus.c@1385 Failed to look up test object for umdns
Wed Apr 24 16:53:24 2024 daemon.err dawn: ubus_call_umdns()=ubus.c@1385 Failed to look up test object for umdns
Wed Apr 24 16:53:40 2024 daemon.err dawn: ubus_call_umdns()=ubus.c@1385 Failed to look up test object for umdns
Wed Apr 24 16:53:56 2024 daemon.err dawn: ubus_call_umdns()=ubus.c@1385 Failed to look up test object for umdns
Wed Apr 24 16:54:12 2024 daemon.err dawn: ubus_call_umdns()=ubus.c@1385 Failed to look up test object for umdns
Wed Apr 24 16:54:28 2024 daemon.err dawn: ubus_call_umdns()=ubus.c@1385 Failed to look up test object for umdns
Wed Apr 24 16:54:44 2024 daemon.err dawn: ubus_call_umdns()=ubus.c@1385 Failed to look up test object for umdns
Wed Apr 24 16:55:00 2024 daemon.err dawn: ubus_call_umdns()=ubus.c@1385 Failed to look up test object for umdns
Wed Apr 24 16:55:16 2024 daemon.err dawn: ubus_call_umdns()=ubus.c@1385 Failed to look up test object for umdns

Background:

I deployed the location newyorck a few weeks ago, with wifi_roaming enabled in the config. It seemed to work fine at first. But after some time I got reports from people not able to connect anymore, sudden disconnects, very changeable reception. When I made the research I found the above and made a pr to remove this and other features from the setup. The errors indicate a problem with the roaming. Anyway I also disabled owe transition mode, just to be sure to have a simple/robust setup. Since then it is working fine.