Proxmox 8 kept dying with error 'UBSAN: array-index-out-of-bounds'

[brok88]

I bought a new Beelink SER6 Mini PC, AMD Ryzen 7 7735HS (8C/16T,up to 4.75GHz) with 16GB ram and had installed Proxmox 8 recently. It was working fine until 3 weeks ago when I realised I am not able to PING to the host or the VM. There is only 1 Windows 10 VM by the way.

I switched off the Beelink and powered it on again, I was able to login to the Proxmox console. However within 5 minutes, it crashed again. When I looked at the monitor, it is showing lines of errors involving UBSAN. When it crashed, I am unable to ping to the host.

I had worked for 20 years involving the Windows and VMWare infrastructures and 1st time trying out Proxmox after hearing how great it was. I tried googled online for the cause and solution but couldn't find anything that is useful for me. This crashing had been going on for a few weeks now and I am beginning to think if Proxmox is really suitable in a Production environment.

If any gurus here can help me look at the errors and point me to the right direction, it will be very helpful for me

 

[Szymons]

I have UBSAN messages only on bootup but this is more megaraid

UBSAN: array-index-out-of-bounds in drivers/scsi/megaraid/megaraid_sas_fp.c:103:32
index 1 is out of range for type 'MR_LD_SPAN_MAP [1]'

https://forum.proxmox.com/threads/proxmox-ve-7-2-megaraid-issues.110587/

The messages are just warnings of a sanitizer - and should not impact the functioning of the system (please correct me if I'm wrong)

about Your issue ... I am not sure but You can try :
- check problems with /etc/default/grub options with 'amd_iommu'

GRUB_CMDLINE_LINUX_DEFAULT="amd_iommu=on iommu=pt quiet"

- check also SMT (HT on amd) and if turned off You must black list ACPI_PAD

##cat /etc/modprobe.d/blacklist.conf
blacklist acpi_pad

- cpuidle_state... check cpu mode :

cat /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor

good luck

 

[Szymons]

ofc after editing grub... run :
update-grub

about cpu idle states ... I am using linux-cpupower and set performance during bootup

/usr/bin/cpupower -c all frequency-set -g performance

or add
/lib/systemd/system/x86_energy_perf_policy.service

[Unit]
Description=Run x86_energy_perf_policy at boot
After=network-online.target

[Service]
Type=oneshot
ExecStart=/usr/bin/cpupower -c all frequency-set -g performance

[Install]
WantedBy=multi-user.target

and
systemctl enable x86_energy_perf_policy.service