Recurring Kernel Panics - Fatal trap 12: page fault while in kernel mode

[rafaelreisr]

Troubleshooting update:

TLDR: Potential solution found - use kvm64 cpu with aes flag enabled.


I have tried different CPU settings on Proxmox:

kvm64: -mitigation flags +aes = stable Only setup with more than 20hs uptime. I reached 48hs.

Code Select Expand

2 (1 sockets, 2 cores) [kvm64,flags=-pcid;-spec-ctrl;-ssbd;-ibpb;-virt-ssbd;-amd-ssbd;-amd-no-ssb;+aes]

Note: kvm64 is a legacy very old Pentium Based CPU with very little flags. Adding AES helps a lot. I didn't see apreciable CPU performance loss on 2.5gbit loads as compared to host-passthrough. NICs are still passthrough and also very stable.. Mitigations are enabled on host kernel.

qemu64: -mitigation flags +aes = unstable Crashes in sub 20hs as usual.
Crashes the same way as cpu passthrough or host-model

other cpu models I tried a few without success. Considering that qemu64, which is a very migration safe CPU, has crashed, I won't bother trying to cherry-pick which flag is causing the issue.

@yourfriendarmando: I believe it is less a hardware issue than a poor BIOS development. I'm sure the work put in by name brand solutions if far more refined. This CPU simplification workaround is a nice find, especially for users in 3rd world countries where importing is extremely expensive and chinese solutions such as these are a decent bang for the buck, although requires work.

@Nearly9892 and other repliers, considering this find, I don't think I'll bother with Proxmox. It would be worth it if I was clusterizing. But for this homelab single deployment it looks overkill.

Next step: I'll go back to the ubuntu ssd and replicate vm settings there, check if it remains as stable as in proxmox - it should, considering the underlying hypervisor is the same, and so far crashing behaviour between proxmox and ubuntu +kvm has been identical.

[sdf_iain]

Sorry to resurrect this thread, but how did you make out?

I'm having the same issue with nearly the same model (n6005 is the only difference). Did you try a more dependable power supply?

[LiFE1688]

Sadly, this is my solution for a N5105/N6005 regarding stability issues. (See attached picture)
Only the J4125 units does not require a fan over it.

CPU set to host
Machine set to q35
Memory allocated 8gb (Ballooning = Off)
PCI Device passthrough for LAN1 ~ LAN4 (6 port units)
PCI Device passthrough for LAN1 ~ LAN3 (4 port units)

The units for 6 ports are using Intel i226
The units for 4 ports are either using Intel i226 or RealTEK RTL8125

It was running for weeks on 22.7.8 then I upgraded.

OPNsense 22.7.10_2 seems to have issues with all my units, J4125 / N5105 / N6005 tested, but only as a VM in Proxmox. I have not tested bare metal. The N5105 / N6005 has stability issues outside OPNsense, when stress testing with Prime95. Memtest86 passes multiple times. I put a USB 140mm fan over the unit, and stability issues with stress testing goes away. OPNsense 22.7.10_2 still has issues crashing (VM in Proxmox), although it takes days to crash. No issues yet after 2 weeks, with pfSesnse 2.7 but that is a  development version.

Sorry to bring up pfSense if that annoys some people. I don't really care what software solutions is being used, as long as it is suitable for the use case.

[Nearly9892]

I went ahead and purchased the AMD 5825U version with 4 i226. It has been perfectly stable with proxmox along side other vms. I had issues with intel 5105 regardless of tweaks. I recommend paying extra for the AMD even if just for piece of mind.

[ProximusAl]

I've been running OPNSense on the N5105 version you have on bare metal for 5 months now with zero issues at all.

Temperature of the device is a steady 43 degrees C.

I know this doesn't help, but wanted to give some input running on Bare Metal.

[misery]

Most help will you find here: https://forum.proxmox.com/threads/vm-freezes-irregularly.111494/
or here: https://forums.servethehome.com/index.php?threads/topton-jasper-lake-quad-i225v-mini-pc-report.36699/page-111

My system is stable after upgrade to Kernel 6.1 and Microcode-Update. C-States are still enabled.

[sdf_iain]

After some LinuxVMs crashes (running k3s and microk8s without any pods or configuration) I decided that Proxmox was the issue and switched to VMware.

My plan is to try that, then Hyper-V, then give up and run bare metal. Something should eventually be stable  ;)

[LiFE1688]

I got my CWWK N6005 working in Proxmox for 6 days without any crashing so far.

Opt-In Kernel 6.1

Code Select Expand


apt update
apt install pve-kernel-6.1


Update Intel Processor Microcode
Edit /etc/apt/sources.list add non-free to the following

Code Select Expand


deb http://ftp.debian.org/debian bullseye main contrib non-free
deb http://ftp.debian.org/debian bullseye-updates main contrib non-free
deb http://security.debian.org bullseye-security main contrib non-free

then

Code Select Expand


apt update
apt install intel-microcode

You can choose to remove non-free from the sources.list file, it doesn't really matter.

Disable CSTATE in GRUB (I am using UEFI with PCIe Passthru) do it according to your install

Code Select Expand


GRUB_CMDLINE_LINUX_DEFAULT="quiet intel_iommu=on iommu=pt intel_idle.max_cstate=1 processor.max_cstate=1"

then

Code Select Expand


update-grub


I do not know what combination of the above is helping, I will see when I have time to re-test it if the VM doesn't crash after at least 10 days.