[BUG] - PVE won't boot properly

[onepamopa]

Hello,

I'm running the following configuration: ASUS Prime TRX40-PRO (BIOS Version 0702 2019/12/17 - latest available), Threadripper 3960x, 32 Gig ddr4, 1TB NVME, nVidia 1070ti.

Last night I tried to install PVE 6.1-1 - as soon as I select install (or debug install with quiet disabled) - PVE installer hangs on the following place (see attachment).
The machine is not frozen - if I plug / unplug USB devices - the action appears on the monitor.

Today I remembered I had an old HDD with 6.0-7 ( 5.0.21-3 kernel ) already installed, so I plugged it into the PC. It booted normally, I had to adjust a few things in /etc/network/interfaces to get network working and after this - I decided to update to the latest version via the GUI.

So, I checked for updates, installed, rebooted - and now - it again gets stuck on the same place as is on the attached image.

Can someone at PVE check what got changed between 6.0-7 and the latest available update, that can cause this?

I'm also attaching 4 update screen shots from the GUI, showing exactly what was working and to what it was upgraded to - and stopped working.

 

[Stefan_R]

It could be that you're running into this bug: https://www.phoronix.com/scan.php?page=news_item&px=Linux-Boot-Threadripper-Zen2MCE

Can you try going into the GRUB menu and adding 'mce=off' to your kernel commandline at the end?

 

[onepamopa]

Thank you, @Stefan_R that fixed it.

 

[Stefan_R]

FYI, to make it persistent you can add it to your GRUB config (/etc/default/grub and run update-grub, or /etc/kernel/cmdline with pve-efiboot-tool refresh if you're using systemd-boot).

There is a kernel patch which allows you to avoid having to completely disable the MCE subsystem, but AFAICT it's not currently staged for inclusion in 5.3, both upstream and in the Ubuntu repos (which we use for our PVE kernel). I'll look into patching our kernel, so you might be able to drop the parameter in the future.

 

[Stefan_R]

And just for the record, pve-kernel includes the fix since version 5.3.13-2, so you should be able to drop the mce=off parameter now.

 

[onepamopa]

I'll try that tomorrow and let you know of the result.

 

[onepamopa]

All is well with mce=off removed. Thank you, @Stefan_R