Kernel crash on OOM on 5.15, 6.1, not 6.6 about openwrt HOT 4 OPEN

Looks like netifd was killed.
echo -17 > /proc/<pid of netifd>/oom_adj
Also will help to disable stress oom_adj manipulation.

from openwrt.

What I suspect happens - init (procd, pid 0) gets killed and system reboots with timer.
I think this makes it stay and restart processes, you may need whole process chain leading to your logs "protected"
echo -17 > /proc/0/oom_adj

from openwrt.

Thanks for hypotheses!

like netifd was killed

I doubt that as kill -9 $(pidof netifd) does not make SSH session freeze.

disable stress oom_adj manipulation

As far as I can see, stress does not tune oom_*_adj variables, at least /proc/$(pidof stress)/oom_*_adj is zero.

init (procd, pid 0) gets killed

I doubt that as well as OOM killer does not usually kill pid=1 according to oom_kill.c.

echo -17 > /proc/1/oom_adj does not help as a workaround, stress still triggers the issue.

from openwrt.

You need to exclude everything involved in your logging chain, procd is supposed to restart rest if those are killed. Probably zram-swap can let your task finish

from openwrt.