Well, if a client is connected and with a still valid dhcp lease, once the "crash" has occurred, ipv4 communications fail.
Likewise if the client has a static ip, ipv4 comms fail on the crash.
Very specifically, if a client set up for dhcp, but does not have a valid lease, AFTER a crash, if this client connects to wireless, it usually does get an ipv4 lease, yet ipv4 comms still fail.
A conclusion that I can come to is that layer 2 comms is still working, so dhcp still works (as it is a layer 2 protocol), but layer 3 is partially broken, preventing ipv4 traffic.
Interestingly, ipv6 traffic seems to still work.... hence my "partially broken" comment.
Only a reboot seems to fix it.
This to me looks like an on-chip firmware crash (in this case on the MediaTek MT7981). A bug, or overheating could both cause this. Just my (currently) speculative view.
I have a MT3000 and I installed the fan and temp sensor.
The fan is alwas on; at its lowest it is 1,500 rpms but I've seen it ramp up to 2,200 and it does it abruptly. 5 second pollig and it goes from 1,500 to 2,000 in those 5.
I do not know what the temp is measuring but it stays between 44.4c and 45c.
I enabled the leds on just kernal hertbeat (white) and the other (blue) is rx and tx.
I think you should get the monitors and enable the leds to see if it is trying to tx and has a heartbeat when it locks up on you.
I did have two lock ups today, both lasted ~3 minutes; the lan switch port went down but I could see the router's leds flickering.
I only use wifi for phones and the like, anything I can put ethernet in I do.
Granted: I don't really tax the thing, I just stream.
It does not lock up, I can still ssh into it using ipv6. It is just ipv4 that stops working. On the crash, from memory, the fan was pulsing once per second-ish - not 5 seconds and the case was very warm compared to normal - not steamy hot, but not "normal" either. This happens on ALL the mt3000s I have, so is not a faulty unit problem.
But yes, good idea re fan and temp sensor stuff.
I just powered one up and about to look for the gpio controls for the fan to see if I can hard code it to full on....
@crcollver@brada4@psherman@LilRedDog
I noticed the mt3000 I am testing on got to the stage of hot/pulsing-fan, with the antennas folded down in the "storage" position.
Lifting them up allowed a stream of warm air to flow out of the left side (left side when the ethernet ports are facing you). Within a few seconds the fan stabilised to a constant fairly high speed... purely measured using my ears and nose so far
Because I am getting the behavior described:
low on temp for the fan, no sudden on and then offs but I do not see stepping fan speed; as the graph shows it is a steady climb.
Squashfs is the only option. The mt3000 has 256MB NAND and 512MB RAM.
But yes, 23.05.3 on this one but have tried snapshot earlier, in approx March 2024.
I flashed one with today's snapshot and it seems much better. The fan at idle is noticeably faster and folding down the antennas to get it to heat up did not make the fan frantically start pulsing, so I need to do a few proper tests on this.
Meanwhile this second mt3000 is now also in the mesh network.
~4 hours and still going. Nothing proven yet though.