On Wed 01-08-18 14:51:25, David Rientjes wrote: > On Tue, 31 Jul 2018, Roman Gushchin wrote: > > > > What's the plan with the cgroup aware oom killer? It has been sitting in > > > the -mm tree for ages with no clear path to being merged. > > > > It's because your nack, isn't it? > > Everybody else seem to be fine with it. > > > > If they are fine with it, I'm not sure they have tested it :) Killing > entire cgroups needlessly for mempolicy oom kills that will not free > memory on target nodes is the first regression they may notice. I do not remember you would be mentioning this previously. Anyway the older implementation has considered the nodemask in memcg_oom_badness. You are right that a cpuset allocation could needlessly select a memcg with small or no memory from the target nodemask which is something I could have noticed during the review. If only I didn't have to spend all my energy to go through repetitive arguments of yours. Anyway this would be quite trivial to resolve in the same function by checking node_isset(node, current->mems_allowed). Thanks for your productive feedback again. Skipping the rest which is yet again repeating same arguments and it doesn't add anything new to the table. -- Michal Hocko SUSE Labs
