On Thu 05-07-18 16:46:21, Andrew Morton wrote:

On Thu, 21 Jun 2018 14:35:20 -0700 (PDT) David Rientjes [off-list ref] wrote:

quoted

The oom reaper ensures forward progress by setting MMF_OOM_SKIP itself if
it cannot reap an mm.  This can happen for a variety of reasons,
including:

 - the inability to grab mm->mmap_sem in a sufficient amount of time,

 - when the mm has blockable mmu notifiers that could cause the oom reaper
   to stall indefinitely,

but we can also add a third when the oom reaper can "reap" an mm but doing
so is unlikely to free any amount of memory:

 - when the mm's memory is mostly mlocked.

Michal has been talking about making the oom-reaper handle mlocked
memory.  Where are we at with that?

I didn't get to mlocked memory yet because blockable mmu notifiers are
more important. And I've already posted patch for that and it is under
discussion [1]. Mlocked memory is next. 
 
[1] http://lkml.kernel.org/r/20180627074421.GF32348@dhcp22.suse.cz

Btw. I still hate this patch and making any timeout user defineable. It
is a wrong approach and my nack to this patch still applies.

-- 
Michal Hocko
SUSE Labs