WiFi LAN crashes on Arch to Arch SSH & File Transfers

dgs · 2022-08-24 17:11:57

Hello all,

I am facing a very weird issue with my home WiFi network and I have yet to figure out how to resolve it. Some guidance would be greatly appreciated.

My set up is rather simple. I have two machines: my laptop I use as my daily driver, an Asus M16 2022 (A) and an older MSI GT72 (B) which I have set up as a file server for my home (samba as there are other non Linux devices which will use it), but both have dual boot with Windows 11 and 10 respectively, and both are running Arch Linux on the latest kernel (5.19.3 as of the time of writing). The router is placed upstairs and I unfortunately do not have access to Ethernet - I rent a basement flat which includes the internet so I only have access to the WiFi(un-metered).

The problem happens most often when I have both systems connected to the 5Ghz network, and seems to happen less often when I have one of them on the 2.4Ghz while keeping the other on the 5Ghz. The main symptom is that as soon as I attempt to ssh into the other machine (it doesn't matter which direction, it happens from either A to B or B to A), the network will crash, meaning I won't be able to browse any web site, ping any other device on the network or anything outside the network, from either of the two machines and often not even on my phone. This symptom is persistent until I disconnect from the WiFi on both, let them idle for a minute or two, and reconnect. Sometimes this doesn't fix the issue however as the network speeds for both pings and browsing will then be minimal (normally I can achieve 170Mbps~200Mbps on a speedtest, where as when this issue happens it'll drop down to 100Kbs). Other times I have to reboot the router (from the web interface but I have asked my upstairs neighbors to do it manually before which didn't make a difference) for things to go back to normal. This same symptom occurs when I try to transfer something to or from the samba shares too. Transferring files through rsync or scp also yield the same results.

The curious thing is that this only happens when either of the two Arch machines are involved, and it's even worse when both are involved - things will crash much sooner, in many cases not even an initial SSH connection will succeed. I have confirmed this because I have attempted to SSH into both of these machines and transfer files from the Windows partitions of each, to and from the Arch samba share, as well as from my phone and from the server's Windows partition onto my daily driver Arch Linux install via scp. The transfers would start at an okay speed but will quickly drop. I have tried using rsync and scp for the file transfers with the same results.

With that said, this all seems to point to a network bottleneck or WiFi interference which would make perfect sense, but then my tests with Linux Mint live mediums seem to say otherwise. It is extremely weird that none of this happens at all through Linux Mint.
For testing purposes, I booted both machines from Linux Mint boot mediums. I then installed openssh-server on them, and set up a samba share on each. I have all my drives connected on both machines so I mounted them to have files and drive space to transfer between them and to my surprises, the network did not crash at all. I tried stressing the network as much as I possibly could, by setting several file transfers at once while spawning many different speed tests concurrently, as well as attempting to SSH from one machine to the other and viceversa. I tried many different files - from small .NEF (Nikon raw files) of 30Mb each, to larger .mov files of 1~2Gb each. I also made some 5 and 10Gb files with 'dd if=/dev/zero of=5gb.img bs=1 count=0 seek=5G', and 10 respectively. I then set all these files to transfer simultaneously through several different terminal windows via scp from both machine A to B and from B to A and also onto the samba shares. I had Firefox opened on both as well during the transfer, from which I kept opening lots of tabs at 'fast.com' to have lots of speed tests done concurrently on both machines. I noticed no out of the ordinary issues with the transfers except some slow downs but they seemed to have divided the speeds evenly and more importantly, they all completed without any drops whatsoever. At the same time, the speed tests were giving very decent results, at aroun d50~60Mbps despite the huge file transfer loads, unlike the most definitely unusable 100Kbs I get throughout all devices by simply attempting to SSH from one Arch machine to the other, or from my phone to either of the Arch systems, or from one Windows partition to the other Arch system.

So this experiment leaves me clueless as to what the culprit could be other than some form of configuration issue with the Arch Linux network configurations on either or both machines. They currently both run NetworkManager, but have tried other network managers such as connman and iwd, on both systems and not at the same time. I have tested using default configurations for mostly everything. The only configurations I have changed has been sshd, for private key authorization but nothing more. I am completely lost as to what else the culprit could be.

Some guidance and ideas I could perhaps get into where to proceed next in terms of troubleshooting this would be greatly appreciated. I did come across this other thread which seems to be of a relatively similar issue, but nothing there seemed to have helped my cause and up to date it was never marked as solved. I also didn't want to post there as I know necrobumping is against the rules. But here is the link as it may help further explain the issue: https://bbs.archlinux.org/viewtopic.php?id=248294&p=2

Thanks in advance!

ewaller · 2022-08-24 17:30:53

Bizarre. No clue.

Static or dynamic addresses? Are the machines getting IPv4 addresses? How about IPv6? Are they local addresses, or do you also get externally routable addresses?
No address collisions?
What if you disable IPv6?

Are only the Arch boxes affected when the problem occurs or does it bring down the net for every machine on the network (sorry if you mentioned that and I missed it)

seth · 2022-08-24 21:08:56

the network will crash, meaning I won't be able to browse any web site, ping any other device on the network or anything outside the network, from either of the two machines and often not even on my phone. This symptom is persistent until I disconnect from the WiFi on both, let them idle for a minute or two, and reconnect. Sometimes this doesn't fix the issue however as the network speeds for both pings and browsing will then be minimal (normally I can achieve 170Mbps~200Mbps on a speedtest, where as when this issue happens it'll drop down to 100Kbs). Other times I have to reboot the router (from the web interface but I have asked my upstairs neighbors to do it manually before which didn't make a difference) for things to go back to normal.

both have dual boot with Windows 11 and 10 respectively

3rd link below, mandatory.
It's not the BIOS setting and windows keeps re-enabling it. Reboot windows and linux and the router afterwards.

Arch Linux

#1 2022-08-24 17:11:57

WiFi LAN crashes on Arch to Arch SSH & File Transfers

#2 2022-08-24 17:30:53

Re: WiFi LAN crashes on Arch to Arch SSH & File Transfers

#3 2022-08-24 21:08:56

Re: WiFi LAN crashes on Arch to Arch SSH & File Transfers

Board footer