On Mon, Oct 02, 2023 at 04:58:21PM +0200, Michal Hocko wrote: > Also there is not OOM as hugetlb pages are costly requests and we do not > invoke the oom killer. Ah good point. That seems like a policy choice we could make. However, since hugetlb users are already set up for and come to expect SIGBUS for physical failure as well as hugetlb_cgroup limits, we should have memcg follow established precedent and leave the OOM killer out. Agree that a sentence in the changelog about this makes sense though.
