Comment 17 for bug 1813244

Andrea Righi (arighi) wrote :

I did more tests with this, it looks like the problem is hidden somewhere in openvswitch. The specific test case that seems to trigger the bug 100% of the time is the one attached, but it requires to change the time immediately before running it (for example using date --set "23:57:00", as suggested before).

An interesting thing that I've tried is to always return an error from ovs_vport_alloc(), this obviously prevents openvswitch from working, but with this change applied I wasn't able to trigger the bug. This requires more investigation, but it looks like somewhere in the openvswitch is corrupting some allocated memory, triggering the kernel bug from other random places.