[SOLVED] SSH over wifi stops working on RT3200/E8450 with 22.03.0-rc6

Hello everyone when you say frozen ssh session it is via putty in ssh only or if I have a computer connected with my belkin rt3200 on rc6 it can freeze or disconnect from the router, for now my devices do not disconnect, touch wood

Ps sorry if I am off topic :relaxed:

It's the former: I had connected into the router with SSH, and was running a program (htop) continuously. See my next note...

3 Likes

Summary of problem and workarounds

When connected to my 22.03.0-rc6 router via Wi-Fi, I would see an SSH connection freeze after a while (minutes to perhaps an hour). I would run htop, and the output from the router would freeze. Furthermore, I could no longer log into the router, even from another SSH session over Wi-FI.

I was surprised to see that the router would continue to pass traffic from my computer to the internet, and I could continue to use the LuCI web interface even when the SSH sessions were hung up. And SSH connections via Ethernet worked just fine(!)

The symptoms seem to be caused by a combination of a bug in the MT76 driver and DSCP markings ("af21") recently introduced by dropbear. This group has proposed workarounds (#1 and #2), both of which seem to address the problem, but the real fix seems to require the MT76 driver update.

Another workaround seems to be simply turning Wi-Fi off and back on on my MacBook Pro - that immediately allowed me to SSH back in.

Update: The workarounds are not the fix - see nbd's patch for what appears to be the real fix

2 Likes

What about windows? Does turning wifi off and on windows too fixed the issue?

I don't know - I didn't test it. Can you detect the problem using a Windows computer? Thanks.

Yes, it happens on Windows using an SSH client.
Yes. it happens on Linux using an SSH client.
Yes, it happens on Android using an SSH client.

We know where the problem is, and it's not the client.

2 Likes

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

I pushed a fix for this to master and 22.03

7 Likes

Thanks heaps for this, @nbd! Building master, as I type, will let this thread know how it goes.

Please, and if you have a moment, can you accept the next PR https://github.com/openwrt/openwrt/pull/10395

1 Like

Awesome, thank you very much. I imagine this will prevent a lot of headache when 22.03 releases.

Which one(s) of the individual commits is/are the relevant one(s)? It's not immediately obvious from the commit titles.

Will this mean that SSH works for me again over 2.4GHz and not only on 5GHz (WDS)?

2 Likes

That commit is for mt7615. Would be weird if it fixes things elsewhere.

edit: it was mentioned the code is also used in mt7622. interesting...

I would like to confirm that in my case it works. Of course, my devices are 2x NanoHD with mt7615E chipset.

2 Likes

Likewise, my slow/unusable ssh on 2.4GHz over WDS issue appears to have been resolved.

1 Like

I installed the snapshot from last night on my Belking RT3200: Powered by LuCI Master (git-22.213.35850-abd9125) / OpenWrt SNAPSHOT r20389-09b086eeca. No additional packages beyond nano and htop; no configuration beyond 2.4GHz wireless and LAN subnet.

Although you cannot prove a negative, the newest snapshot has been running for 3+ hours (best previous record was ~1 hour) I remain hopeful...

Update: after 6 hours 9 hours 20+ hours, still running htop successfully. It's looking pretty good that @nbd's patch is a fix. Thanks to all who helped.

6 Likes

Thanks @richb-hanover-priv I complained about slow ssh on 2.4GHz but didn't get anywhere. Your methodical approach helped effect change. And many thanks indeed to @nbd here for continually making this better and better.

1 Like

Closing this topic. If we see evidence of a new problem, open a new topic