Description: I have 2 10gb Nic's with the Aquantia chipsets, plugged in through a Netgear GS110EMX 10gb switch. One Nic is plugged into a windows 10 host and the other is in the unraid server. When ever I plug the windows PC into the switch the Unraid server loses connectivity and has an NMI error on CPU0. The error is also known as the Dazed and confused error. The machine has to be rebooted in order to get connectivity and the pc needs to be unplugged in order the machine to get online again.

 

How to reproduce: Plug another 10gb device into the same network at unraid.

 

Expected results: 10gb connectivity between unraid and the windows pc.

 

Actual results: Unraid loses all network connectivity, including other network cards connected to the server.

 

Other information: This seems to be a bug related to the linux driver implemented for the Aquantia chipsets.

 

Exact error:  "NMI: PCI system error (SERR) for reason b1 on CPU0. Dazed and confused, but trying to continue"

 

mightybox-diagnostics-20180113-1859.zip

Edited January 14, 20188 yr by rudeytwotone
Upgraded to latest stable and rebuilt diagnostic package

RC is probably no longer supported since 6.4 is not "stable." You should upgrade and then fix the title to reflect that.

Thanks, updated and also updated diagnostic info. unfortunately still happening. 

Hi Alex,

 

We've looked into this issue and have determined this may be the result of a buggy driver implementation in the Linux kernel related to this device.  In your e-mail to us, you indicated ASUS has provided you with an out-of-tree driver that doesn't match what is in the Linux kernel that we use.  I would encourage you to contact ASUS and urge them to consider implementing this supposedly "better" driver in the kernel tree if it will result in better experiences for their customers.

 

In short, we will not be building in this out-of-tree driver for your device into unRAID.  The reason is quite simple:  it could break things for others.  Previously we implemented an out-of-tree Linux driver for Intel's 10gbe NICs and while some folks reported better experiences after this, others reported numerous problems.  Later we reverted to the in-tree driver in a later kernel revision and everyone was happy.  The issue here is that we don't know what building this out-of-tree driver might do and we can't guarantee that it will always build correctly against future kernels.  So in the meantime, the recommended solution for using 10gbe on unRAID is to procure Intel-based 10gbe NICs (or at least NICs that have been tested and vetted here by the unRAID community).  I wish I had a better answer for you, but the hard truth is that ASUS may just not have as much love for Linux as they do for Windows.