On Thu, Jan 13, 2022 at 02:08:10PM +0100, Sebastian Andrzej Siewior <bigeasy@xxxxxxxxxxxxx> wrote: > I added a preempt-disable() section restricted to RT to > mem_cgroup_event_ratelimit(). Oh, I missed that one. (Than the decoupling of such mem_cgroup_event_ratelimit() also makes some more sense.) > That would mean that mem_cgroup_event_ratelimit() needs a > local_irq_save(). If that is okay then sure I can move it that way. Whatever avoids the twisted code :-) --- > I remember Michal (Hocko) suggested excluding/ rejecting soft limit but > I didn't know where exactly and its implications. In this block here I > just followed the replacement of irq-off with preempt-off for RT. Both soft limit and (these) event notifications are v1 features. Soft limit itself is rather considered even misfeature. I guess the implications would not be many since PREEMPT_RT+memcg users would be new(?) so should rather start with v2 anyway. One way to disable it would be to reject writes into memory.soft_limit_in_bytes or cgroup.event_control + documentation of that. Michal
