On Fri, 17 Apr 2020 09:37:16 +0200 Michal Hocko wrote: > On Thu 16-04-20 18:06:17, Jakub Kicinski wrote: > > Add a memory.swap.high knob, which functions in a similar way > > to memory.high, but with a less steep slope. > > This really begs for more details. What does "similar way to > memory.high" mean? Because my first thought would be that the swap usage > will be throttled and kept at the swap.high level. From a quick look at > the patch you only do throttling. There is no swap space reclaim to keep > the usage at the level. So unless I am missing something this is very > confusing and it doesn't really fit into high limit semantic. The new knob does influence mem_cgroup_swap_full(), but you're right, the analogy is not strong enough, I'll rephrase. > This changelog also doesn't describe any usecase. Will do.
