Comment 12 for bug 1837810

As requested by the kernel team (in https://lists.ubuntu.com/archives/kernel-team/2020-August/112775.html), I will do some additional testing for this SRU to really make sure it won't cause any regressions.

I provisioned a lab machine on segmaas, running Bionic. I installed the 4.15.0-116-generic kernel from -proposed on it.

I built the zero_page_refcount.c kernel module, and inserted it into the running kernel.

I then got ksm_refcnt_overflow.sh running in a screen session, creating and destroying virtual machines in an infinite loop.

This way we will know the code path has been exercised a fair amount.

I will leave this running creating and destroying virtual machines for a week or so, and I will report back with the results.