From: Vlastimil Babka <vbabka@suse.cz>
To: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Andrew Morton <akpm@linux-foundation.org>,
linux-mm@kvack.org, linux-kernel@vger.kernel.org,
Minchan Kim <minchan@kernel.org>, Mel Gorman <mgorman@suse.de>,
Michal Nazarewicz <mina86@mina86.com>,
Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>,
Christoph Lameter <cl@linux.com>, Rik van Riel <riel@redhat.com>,
David Rientjes <rientjes@google.com>
Subject: Re: [PATCH 5/5] mm, compaction: more focused lru and pcplists draining
Date: Mon, 03 Nov 2014 09:12:33 +0100 [thread overview]
Message-ID: <545738F1.4010307@suse.cz> (raw)
In-Reply-To: <20141027074112.GC23379@js1304-P5Q-DELUXE>
On 10/27/2014 08:41 AM, Joonsoo Kim wrote:
> On Tue, Oct 07, 2014 at 05:33:39PM +0200, Vlastimil Babka wrote:
>> The goal of memory compaction is to create high-order freepages through page
>> migration. Page migration however puts pages on the per-cpu lru_add cache,
>> which is later flushed to per-cpu pcplists, and only after pcplists are
>> drained the pages can actually merge. This can happen due to the per-cpu
>> caches becoming full through further freeing, or explicitly.
>>
>> During direct compaction, it is useful to do the draining explicitly so that
>> pages merge as soon as possible and compaction can detect success immediately
>> and keep the latency impact at minimum. However the current implementation is
>> far from ideal. Draining is done only in __alloc_pages_direct_compact(),
>> after all zones were already compacted, and the decisions to continue or stop
>> compaction in individual zones was done without the last batch of migrations
>> being merged. It is also missing the draining of lru_add cache before the
>> pcplists.
>>
>> This patch moves the draining for direct compaction into compact_zone(). It
>> adds the missing lru_cache draining and uses the newly introduced single zone
>> pcplists draining to reduce overhead and avoid impact on unrelated zones.
>> Draining is only performed when it can actually lead to merging of a page of
>> desired order (passed by cc->order). This means it is only done when migration
>> occurred in the previously scanned cc->order aligned block(s) and the
>> migration scanner is now pointing to the next cc->order aligned block.
>>
>> The patch has been tested with stress-highalloc benchmark from mmtests.
>> Although overal allocation success rates of the benchmark were not affected,
>> the number of detected compaction successes has doubled. This suggests that
>> allocations were previously successful due to implicit merging caused by
>> background activity, making a later allocation attempt succeed immediately,
>> but not attributing the success to compaction. Since stress-highalloc always
>> tries to allocate almost the whole memory, it cannot show the improvement in
>> its reported success rate metric. However after this patch, compaction should
>> detect success and terminate earlier, reducing the direct compaction latencies
>> in a real scenario.
>>
>> Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
>> Cc: Minchan Kim <minchan@kernel.org>
>> Cc: Mel Gorman <mgorman@suse.de>
>> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
>> Cc: Michal Nazarewicz <mina86@mina86.com>
>> Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
>> Cc: Christoph Lameter <cl@linux.com>
>> Cc: Rik van Riel <riel@redhat.com>
>> Cc: David Rientjes <rientjes@google.com>
>> ---
>> mm/compaction.c | 41 ++++++++++++++++++++++++++++++++++++++++-
>> mm/page_alloc.c | 4 ----
>> 2 files changed, 40 insertions(+), 5 deletions(-)
>>
>> diff --git a/mm/compaction.c b/mm/compaction.c
>> index 8fa888d..41b49d7 100644
>> --- a/mm/compaction.c
>> +++ b/mm/compaction.c
>> @@ -1179,6 +1179,7 @@ static int compact_zone(struct zone *zone, struct compact_control *cc)
>> while ((ret = compact_finished(zone, cc, migratetype)) ==
>> COMPACT_CONTINUE) {
>> int err;
>> + unsigned long last_migrated_pfn = 0;
>
> I think that this definition looks odd.
> In every iteration, last_migrated_pfn is re-defined as 0.
> Maybe, it is on outside of the loop.
Oops you're right, that's a mistake and it makes the code miss some of
the drain points (a minority I think but anyway).
>>
>> switch (isolate_migratepages(zone, cc)) {
>> case ISOLATE_ABORT:
>> @@ -1187,7 +1188,12 @@ static int compact_zone(struct zone *zone, struct compact_control *cc)
>> cc->nr_migratepages = 0;
>> goto out;
>> case ISOLATE_NONE:
>> - continue;
>> + /*
>> + * We haven't isolated and migrated anything, but
>> + * there might still be unflushed migrations from
>> + * previous cc->order aligned block.
>> + */
>> + goto check_drain;
>> case ISOLATE_SUCCESS:
>> ;
>> }
>> @@ -1212,6 +1218,39 @@ static int compact_zone(struct zone *zone, struct compact_control *cc)
>> goto out;
>> }
>> }
>> +
>> + /*
>> + * Record where we have freed pages by migration and not yet
>> + * flushed them to buddy allocator. Subtract 1, because often
>> + * we finish a pageblock and migrate_pfn points to the first
>> + * page* of the next one. In that case we want the drain below
>> + * to happen immediately.
>> + */
>> + if (!last_migrated_pfn)
>> + last_migrated_pfn = cc->migrate_pfn - 1;
>
> And, I wonder why last_migrated_pfn is set after isolate_migratepages().
Not sure I understand your question. With the mistake above, it cannot
currently be set at the point isolate_migratepages() is called, so you
might question the goto check_drain in the ISOLATE_NONE case, if that's
what you are wondering about.
When I correct that, it might be set when COMPACT_CLUSTER_MAX pages are
isolated and migrated the middle of a pageblock, and then the rest of
the pageblock contains no pages that could be isolated, so the last
isolate_migratepages() attempt in the pageblock returns with
ISOLATE_NONE. Still there were some migrations that produced free pages
that should be drained at that point.
>
> Thanks.
>
next prev parent reply other threads:[~2014-11-03 8:12 UTC|newest]
Thread overview: 30+ messages / expand[flat|nested] mbox.gz Atom feed top
2014-10-07 15:33 [PATCH 0/5] Further compaction tuning Vlastimil Babka
2014-10-07 15:33 ` [PATCH 1/5] mm, compaction: pass classzone_idx and alloc_flags to watermark checking Vlastimil Babka
2014-10-20 15:45 ` Rik van Riel
2014-10-27 6:46 ` Joonsoo Kim
2014-10-27 9:11 ` Vlastimil Babka
2014-10-28 7:16 ` Joonsoo Kim
2014-10-29 13:51 ` Vlastimil Babka
2014-10-31 7:49 ` Joonsoo Kim
2014-11-14 8:52 ` Vlastimil Babka
2014-10-07 15:33 ` [PATCH 2/5] mm, compaction: simplify deferred compaction Vlastimil Babka
2014-10-15 22:32 ` Andrew Morton
2014-10-16 15:11 ` Vlastimil Babka
2014-10-07 15:33 ` [PATCH 3/5] mm, compaction: defer only on COMPACT_COMPLETE Vlastimil Babka
2014-10-20 15:18 ` Rik van Riel
2014-10-07 15:33 ` [PATCH 4/5] mm, compaction: always update cached scanner positions Vlastimil Babka
2014-10-20 15:26 ` Rik van Riel
2014-10-27 7:35 ` Joonsoo Kim
2014-10-27 9:39 ` Vlastimil Babka
2014-10-28 7:08 ` Joonsoo Kim
2014-10-31 15:53 ` Vlastimil Babka
2014-11-04 0:28 ` Joonsoo Kim
2014-11-14 8:57 ` Vlastimil Babka
2014-10-07 15:33 ` [PATCH 5/5] mm, compaction: more focused lru and pcplists draining Vlastimil Babka
2014-10-20 15:44 ` Rik van Riel
2014-10-27 7:41 ` Joonsoo Kim
2014-11-03 8:12 ` Vlastimil Babka [this message]
2014-11-04 0:37 ` Joonsoo Kim
2014-11-13 12:47 ` Vlastimil Babka
2014-11-14 7:05 ` Joonsoo Kim
2014-11-19 22:53 ` Vlastimil Babka
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=545738F1.4010307@suse.cz \
--to=vbabka@suse.cz \
--cc=akpm@linux-foundation.org \
--cc=cl@linux.com \
--cc=iamjoonsoo.kim@lge.com \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-mm@kvack.org \
--cc=mgorman@suse.de \
--cc=mina86@mina86.com \
--cc=minchan@kernel.org \
--cc=n-horiguchi@ah.jp.nec.com \
--cc=riel@redhat.com \
--cc=rientjes@google.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).