Opt in kernel panics

[Kodey]

I'd like to upgrade to 8, but all these 6.2 kernels keep spontaneously rebooting. 5.19 and 5.15.64 is working with gpu passthrough on this amd but I'm not using vendor reset, just claiming the pci devices as vfio-pci.ids with cmdline and masking the modules then resetting the interface on vm startup hooks.
6.2 runs sluggishly then freezes or reboots. No idea why as no messages are logged

 

[t.lamprecht]

I'd like to upgrade to 8, but all these 6.2 kernels keep spontaneously rebooting. 5.19 and 5.15.64 is working with gpu passthrough on this amd but I'm not using vendor reset, just claiming the pci devices as vfio-pci.ids with cmdline and masking the modules then resetting the interface on vm startup hooks.
6.2 runs sluggishly then freezes or reboots. No idea why as no messages are logged

What hardware (CPU, motherboard, GPU/PCIe cards you pass through) are we talking here?

Is the BIOS/Firmware up to date? Sometimes updating those can improve compatibility with newer kernels too.

 

[Kodey]

What hardware (CPU, motherboard, GPU/PCIe cards you pass through) are we talking here?

Is the BIOS/Firmware up to date? Sometimes updating those can improve compatibility with newer kernels too.

Hi Thanks,
Turned out to be caused by hugepages and zfs. I had over committed huge pages in kernel parameters (50% ram) and assigned all and later half of that to one virtual machine because it was so slow to startup without it. Zfs was also consuming a lot of memory in the background which I had limited but to no avail. The machine regularly kept dropping out or hanging randomly until I dropped the hugepages down radically and further limited zfs. Now it's been running solidly for a few weeks.
This was difficult to attribute as those setting had been working fine for a long time with older kernels 5.15, 5.19 when there was less vm and zfs activity. It highlights the need to accurately see how memory is being used by zfs and hugepages.