After updating to 24.10 I’m constantly getting a dozen of these log messages spammed over and over on the console logs for the single network interface:
bad gso: type: 4, size: 8940
I was doing some checking and this seems to be related, in some way, to file transfers to TrueNAS. I’m getting maybe 5MB/s transfer speeds at best and the messages spam whenever I try to transfer files but stop whenever there are no transfers going on.
This seems to be a result of a new Linux kernel bug:
https://bugzilla.kernel.org/show_bug.cgi?id=219129
Edit:
For now it seems Intel E1000E works fine but vxmnet3 and VirtIO have issues.
Switched to E1000E too as a workaround
Hmm. I have 2 VMs hosted in my box running EE, and I’m not seeing this problem. In reading the bugzilla report, it sounds like it’s specific to jumbo frames potentially. It will likely be less performance impacting to use standard MTU sizes in conjunction with a VIRTIO networking type on any interface greater than 1 Gigabit.
The issue is not present with EE interfaces, only with VirtIO and vxmnet3 interface types. Tried with and without jumbo frames and no change in performance either way with VirtIO or vmxnet3
I have the exact same issue with VirtIO, switched to E1000E as a workaround aswell.