Subject: Re: [PATCH] mm/vmstat: retrieve more accurate vmstat value
To: Michal Hocko <mhocko@kernel.org>, Joonsoo Kim <js1304@gmail.com>
References: <1448346123-2699-1-git-send-email-iamjoonsoo.kim@lge.com>
 <20151125120021.GA27342@dhcp22.suse.cz>
Cc: Andrew Morton <akpm@linux-foundation.org>, linux-kernel@vger.kernel.org,
        Christoph Lameter <cl@linux.com>, linux-mm@kvack.org,
        Joonsoo Kim <iamjoonsoo.kim@lge.com>
From: Vlastimil Babka <vbabka@suse.cz>
Message-ID: <5655BB0A.90000@suse.cz>
Date: Wed, 25 Nov 2015 14:43:38 +0100
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:38.0) Gecko/20100101
 Thunderbird/38.3.0
MIME-Version: 1.0
In-Reply-To: <20151125120021.GA27342@dhcp22.suse.cz>
Content-Type: text/plain; charset=windows-1252
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 1213
Lines: 25

On 11/25/2015 01:00 PM, Michal Hocko wrote:
> On Tue 24-11-15 15:22:03, Joonsoo Kim wrote:
>> When I tested compaction in low memory condition, I found that
>> my benchmark is stuck in congestion_wait() at shrink_inactive_list().
>> This stuck last for 1 sec and after then it can escape. More investigation
>> shows that it is due to stale vmstat value. vmstat is updated every 1 sec
>> so it is stuck for 1 sec.
> 
> Wouldn't it be sufficient to use zone_page_state_snapshot in
> too_many_isolated?

That sounds better than the ad-hoc half-solution, yeah.
I don't know how performance sensitive the callers are, but maybe it could do a
non-snapshot check first, and only repeat with _snapshot when it's about to wait
(the result is true), just to make sure?

OTOH, how big issue is this? I suspect the system has been genuinely
too_many_isolated(), or very close, in order to hit the condition in the first
place, and the inaccuracy just delays the recovery a bit?

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/