Re: [PATCH 3/4] mm: unreserve highatomic free pages fully before OOM
From: Michal Hocko
Date: Tue Oct 11 2016 - 02:52:42 EST
On Tue 11-10-16 14:01:41, Minchan Kim wrote:
> Hi Michal,
> On Mon, Oct 10, 2016 at 09:41:40AM +0200, Michal Hocko wrote:
> > On Fri 07-10-16 23:43:45, Minchan Kim wrote:
> > > On Fri, Oct 07, 2016 at 11:09:17AM +0200, Michal Hocko wrote:
> > > > @@ -2102,10 +2109,12 @@ static void unreserve_highatomic_pageblock(const struct alloc_context *ac)
> > > > set_pageblock_migratetype(page, ac->migratetype);
> > > > move_freepages_block(zone, page, ac->migratetype);
> > > > spin_unlock_irqrestore(&zone->lock, flags);
> > > > - return;
> > > > + return true;
> > >
> > > Such cut-off makes reserved pageblock remained before the OOM.
> > > We call it as premature OOM kill.
> > Not sure I understand. The above should get rid of all atomic reserves
> > before we go OOM. We can do it all at once but that sounds too
> The problem is there is race between page freeing path and unreserve
> logic so that some pages could be in highatomic free list even though
> zone->nr_reserved_highatomic is already zero.
Does it make any sense to handle such an unlikely case?
> So, at least, it would be better to have a draining step at some point
> where was (no_progress_loops == MAX_RECLAIM RETRIES) in my patch.
> Also, your patch makes retry loop greater than MAX_RECLAIM_RETRIES
> if unreserve_highatomic_pageblock returns true. Theoretically,
> it would make live lock. You might argue it's *really really* rare
> but I don't want to add such subtle thing.
> Maybe, we could drain when no_progress_loops == MAX_RECLAIM_RETRIES.
What would be the scenario when we would really livelock here? How can
we have unreserve_highatomic_pageblock returning true for ever?
> > aggressive to me. If we just do one at the time we have a chance to
> > keep some reserves if the OOM situation is really ephemeral.
> > Does this patch work in your usecase?
> I didn't test but I guess it works but it has problems I mentioned
Please do not make this too over complicated and be practical. I do not
really want to dismiss your usecase but I am really not convinced that
such a "perfectly fit into all memory" situations are sustainable and
justify to make the whole code more complex. I agree that we can at
least try to do something to release those reserves but let's do it
as simple as possible.