Re: [PATCH] mm,page_alloc,cma: conditionally prefer cma pageblocks for movable allocations

From: Joonsoo Kim
Date: Fri Mar 13 2020 - 03:44:48 EST


2020ë 3ì 13ì (ê) ìì 2:07, Roman Gushchin <guro@xxxxxx>ëì ìì:
>
> On Thu, Mar 12, 2020 at 05:56:34PM +0900, Joonsoo Kim wrote:
> > 2020ë 3ì 12ì (ë) ìì 11:40, Roman Gushchin <guro@xxxxxx>ëì ìì:
> > >
> > > On Thu, Mar 12, 2020 at 10:41:28AM +0900, Joonsoo Kim wrote:
> > > > Hello, Roman.
> > >
> > > Hello, Joonsoo!
> > >
> > > >
> > > > 2020ë 3ì 12ì (ë) ìì 2:35, Roman Gushchin <guro@xxxxxx>ëì ìì:
> > > > >
> > > > > On Wed, Mar 11, 2020 at 09:51:07AM +0100, Vlastimil Babka wrote:
> > > > > > On 3/6/20 9:01 PM, Rik van Riel wrote:
> > > > > > > Posting this one for Roman so I can deal with any upstream feedback and
> > > > > > > create a v2 if needed, while scratching my head over the next piece of
> > > > > > > this puzzle :)
> > > > > > >
> > > > > > > ---8<---
> > > > > > >
> > > > > > > From: Roman Gushchin <guro@xxxxxx>
> > > > > > >
> > > > > > > Currently a cma area is barely used by the page allocator because
> > > > > > > it's used only as a fallback from movable, however kswapd tries
> > > > > > > hard to make sure that the fallback path isn't used.
> > > > > >
> > > > > > Few years ago Joonsoo wanted to fix these kinds of weird MIGRATE_CMA corner
> > > > > > cases by using ZONE_MOVABLE instead [1]. Unfortunately it was reverted due to
> > > > > > unresolved bugs. Perhaps the idea could be resurrected now?
> > > > >
> > > > > Hi Vlastimil!
> > > > >
> > > > > Thank you for this reminder! I actually looked at it and also asked Joonsoo in private
> > > > > about the state of this patch(set). As I understand, Joonsoo plans to resubmit
> > > > > it later this year.
> > > > >
> > > > > What Rik and I are suggesting seems to be much simpler, however it's perfectly
> > > > > possible that Joonsoo's solution is preferable long-term.
> > > > >
> > > > > So if the proposed patch looks ok for now, I'd suggest to go with it and return
> > > > > to this question once we'll have a new version of ZONE_MOVABLE solution.
> > > >
> > > > Hmm... utilization is not the only matter for CMA user. The more
> > > > important one is
> > > > success guarantee of cma_alloc() and this patch would have a bad impact on it.
> > > >
> > > > A few years ago, I have tested this kind of approach and found that increasing
> > > > utilization increases cma_alloc() failure. Reason is that the page
> > > > allocated with
> > > > __GFP_MOVABLE, especially, by sb_bread(), is sometimes pinned by someone.
> > > >
> > > > Until now, cma memory isn't used much so this problem doesn't occur easily.
> > > > However, with this patch, it would happen.
> > >
> > > Sure, but the whole point of cma is to be able to use the cma area
> > > effectively by movable pages. Otherwise we can just reserve it and
> > > have 100% reliability.
> >
> > I agree with that cma area should be used effectively. However, cma_alloc()
> > failure is functional failure in embedded system so we need to approach
> > this problem more carefully. At least, to control the behaviour, configurable
> > option (default is current behaviour) would be necessary.
>
> Do we know who can test it? Adding a configuration option is a last resort
> measure, I really hope we can avoid it.

I don't know. Agreed that configuration option is a last resort.

> >
> > > So I'd focus on fixing page migration issues, rather than trying
> > > to keep it empty most of the time.
> >
> > Great! Fixing page migration issue is always good thing!
> >
> > > Btw, I've fixed two issues, which dramatically increased the success
> > > rate of 1 GB page allocation in my case:
> > >
> > > https://patchwork.kernel.org/patch/11420997/
> > > https://lore.kernel.org/patchwork/patch/1202868/
> > >
> > > (They both are on the way to upstream, but not there yet)
> > >
> > > Can you, please, pull them and try?
> >
> > Unfortunately, I lose my test setup for this problem so cannot try it now.
> > I will try it as soon as possible.
>
> Thanks! Looking forward to it...
>
> >
> > Anyway, AFAIR, I saw the problem while journal is continually working
> > on ext4. Have you checked this case?
>
> My ext4 fix sounds very similar to what you're describing, but it's hard to
> say for sure.

Okay, I will test it.

Thanks.