Comment 24 for bug 2032176

Revision history for this message
Thadeu Lima de Souza Cascardo (cascardo) wrote :

I have managed to reproduce this by using a large openstack instance (64GiB on an AMD CPU that will use SVM as virtualization technology) and creating an LXD VM with large amounts of memory (32GiB). Then starting and stopping the VM in a loop reproduces the issue. I am running it with tdp_mmu off and so far, I haven't seen it fail.

Given that tdp_mmu may set set as off by default on 5.15.y upstream (see https://<email address hidden>/), this might be the solution for this issue now on our 5.15 trees.

I will test 6.2 with the "reproducer" and we can look into any necessary arrangements on that version as well.

As for 5.19, we don't plan for any updates there anymore.

Cascardo.