On Mon, Aug 06, 2018 at 11:19:26AM -0700, Tejun Heo wrote: > On Mon, Aug 06, 2018 at 11:08:45AM -0700, Andrew Morton wrote: > > On Mon, 6 Aug 2018 09:15:29 -0700 Tejun Heo <tj@xxxxxxxxxx> wrote: > > > > > mem_cgroup_print_oom_info() currently prints the same info for cgroup1 > > > and cgroup2 OOMs. It doesn't make much sense on cgroup2, which > > > doesn't use memsw or separate kmem accounting - the information > > > reported is both superflous and insufficient. This patch updates the > > > memcg OOM messages on cgroup2 so that > > > > > > * It prints memory and swap usages and limits used on cgroup2. > > > > > > * It shows the same information as memory.stat. > > > > > > I took out the recursive printing for cgroup2 because the amount of > > > output could be a lot and the benefits aren't clear. An example dump > > > follows. > > > > This conflicts rather severely with Shakeel's "memcg: reduce memcg tree > > traversals for stats collection". Can we please park this until after > > 4.19-rc1? > > Sure, or I can refresh the patch on top of -mm too. Now that 4.19 is released, do you mind refreshing this for 4.20?
