Review Request: WisdPi USB 3.2 5Gbps Wired Base-T Ethernet Adapter (WP-UT5) based on Realtek RTL8157

[jerrythea]

The problem again is that with the Realtek code, if you host an NFS share (either directly on this adapter, or maybe anywhere) the Proxmox Kernel (either 6.8 or 6.11) will slow down, hang, and then crash.

I spent weeks trying to isolate this, and it is indeed specific to the combination of NFS (ver 3 or ver 4) and this adapter+driver combo.

This behavior was consistent on 5 different systems (all NUC like, 3 11th gen i5, 1 11th i7, 1 Ultra 5 [series 1]).

I have been using these adapters on client systems, even including Fedora 42, without issues.

It is just something to do with an interaction between NFS (kernel based) and this driver.

I don't think I can file a kernel bug, since this code has not yet landed to 6.15++.

 

[WhiteNoise]

Crazy thought. You can find a person with an @realtek.com email address who have contributed to the kernel module and send them an email with your findings.

 

[John T Davis]

Crazy thought. You can find a person with an @realtek.com email address who have contributed to the kernel module and send them an email with your findings.

That's a great idea. Could you help with that?

The problem again is that with the Realtek code, if you host an NFS share (either directly on this adapter, or maybe anywhere) the Proxmox Kernel (either 6.8 or 6.11) will slow down, hang, and then crash.

I spent weeks trying to isolate this, and it is indeed specific to the combination of NFS (ver 3 or ver 4) and this adapter+driver combo.

This behavior was consistent on 5 different systems (all NUC like, 3 11th gen i5, 1 11th i7, 1 Ultra 5 [series 1]).

I have been using these adapters on client systems, even including Fedora 42, without issues.

It is just something to do with an interaction between NFS (kernel based) and this driver.

I don't think I can file a kernel bug, since this code has not yet landed to 6.15++.

@jerrythea, thanks for bringing this up again. I was aware of it, but haven't gotten far enough in setting up the little N100 test Proxmox node I'm using the NIC with to run into it. I've never seen a NIC driver bug tied to a specific protocol like that. I'm sure it's not unique, but it's (luckily for me, I guess) the first time I've run into it.

Since the adapters work on "client systems," is it correct that connecting to an NFS share with Proxmox shouldn't cause a problem? Proxmox with the NIC is the client in that sitaution. Or have I misunderstood something?

Also, the current test kernel for Proxmox is 6.14. Have you tested it with that yet? If not, I can try 6.14 on my test node.

I'd like to report an issue on Realtek's DKMS repo for the driver, here: GitHub - wisdpi/wp-ut5_dkms: A DKMS package for easy use of WP-UT5
That might not accomplish much, but it would potentially warn anyone who grabs that DKMS and doesn't know about the NFS issue (hello, I'm the tester, it's me).

What's the most succinct, helpful, description of bug? I haven't run into this issue yet so I don't have logs or anything. I've seen several descriptions on various threads, but I'm not sure I've seen the most complete diagnostic info.

It might also be worth mentioning on the Proxmox forum to see if anyone there can dig into the driver to figure out a bit more about what's going on.