Re: [PATCH 01/10] mm, page_alloc: Delete the zonelist_cache
From: Mel Gorman
Date: Thu Aug 20 2015 - 10:17:32 EST
On Thu, Aug 20, 2015 at 03:30:54PM +0200, Vlastimil Babka wrote:
> >Note the maximum stall latency which was 6 seconds and becomes 67ms with
> >this patch applied. However, also note that it is not guaranteed this
> >benchmark always hits pathelogical cases and the milage varies. There is
> >a secondary impact with more direct reclaim because zones are now being
> >considered instead of being skipped by zlc.
> >
> > 4.1.0 4.1.0
> > vanilla nozlc-v1r4
> >Swap Ins 838 502
> >Swap Outs 1149395 2622895
> >DMA32 allocs 17839113 15863747
> >Normal allocs 129045707 137847920
> >Direct pages scanned 4070089 29046893
> >Kswapd pages scanned 17147837 17140694
> >Kswapd pages reclaimed 17146691 17139601
> >Direct pages reclaimed 1888879 4886630
> >Kswapd efficiency 99% 99%
> >Kswapd velocity 17523.721 17518.928
> >Direct efficiency 46% 16%
> >Direct velocity 4159.306 29687.854
> >Percentage direct scans 19% 62%
> >Page writes by reclaim 1149395.000 2622895.000
> >Page writes file 0 0
> >Page writes anon 1149395 2622895
>
> Interesting, kswapd has no decrease that would counter the increase in
> direct reclaim. So there's more reclaim overall. Does it mean that stutter
> doesn't like LRU and zlc was disrupting LRU?
>
The LRU is being heavily disrupted by both reclaim and compaction
activity. The test is not a reliable means of evaluating reclaim decisions
because of the compaction activity. The main purpose of stutter was as a
proxy measure of desktop interactivity during IO.
As the test does THP allocations, it can trigger the case where zlc can
disable a zone for no reason and instead busy loop which is just wrong.
> >The direct page scan and reclaim rates are noticeable. It is possible
> >this will not be a universal win on all workloads but cycling through
> >zonelists waiting for zlc->last_full_zap to expire is not the right
> >decision.
> >
> >Signed-off-by: Mel Gorman <mgorman@xxxxxxxxxxxxxxxxxxx>
> >Acked-by: David Rientjes <rientjes@xxxxxxxxxx>
>
> It doesn't seem that removal of zlc would increase overhead due to
> "expensive operations no longer being avoided". Making some corner-case
> benchmark(s) worse as a side-effect of different LRU approximation shouldn't
> be a show-stopper. Hence
>
> Acked-by: Vlastimil Babka <vbabka@xxxxxxx>
>
Thanks.
> just git grep found some lines that should be also deleted:
>
> include/linux/mmzone.h: * If zlcache_ptr is not NULL, then it is just the
> address of zlcache,
> include/linux/mmzone.h: * as explained above. If zlcache_ptr is NULL, there
> is no zlcache.
>
Thanks
> And:
>
> >@@ -3157,7 +2967,7 @@ __alloc_pages_nodemask(gfp_t gfp_mask, unsigned int order,
> > gfp_t alloc_mask; /* The gfp_t that was actually used for allocation */
> > struct alloc_context ac = {
> > .high_zoneidx = gfp_zone(gfp_mask),
> >- .nodemask = nodemask,
> >+ .nodemask = nodemask ? : &cpuset_current_mems_allowed,
> > .migratetype = gfpflags_to_migratetype(gfp_mask),
> > };
> >
> >@@ -3188,8 +2998,7 @@ __alloc_pages_nodemask(gfp_t gfp_mask, unsigned int order,
> > ac.zonelist = zonelist;
> > /* The preferred zone is used for statistics later */
> > preferred_zoneref = first_zones_zonelist(ac.zonelist, ac.high_zoneidx,
> >- ac.nodemask ? : &cpuset_current_mems_allowed,
> >- &ac.preferred_zone);
> >+ ac.nodemask, &ac.preferred_zone);
> > if (!ac.preferred_zone)
> > goto out;
> > ac.classzone_idx = zonelist_zone_idx(preferred_zoneref);
>
> These hunks appear unrelated to zonelist cache? Also they move the
> evaluation of cpuset_current_mems_allowed
They are rebase-related brain damage :(. I'll fix it and retest.
--
Mel Gorman
SUSE Labs
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/