linux-kernel - Re: [PATCH] bdi flusher should not be throttled here when it fall into buddy slow path

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20161018095912.GD22174@techsingularity.net>
Date:   Tue, 18 Oct 2016 10:59:12 +0100
From:   Mel Gorman <mgorman@...hsingularity.net>
To:     zhouxianrong@...wei.com
Cc:     linux-mm@...ck.org, linux-kernel@...r.kernel.org,
        akpm@...ux-foundation.org, viro@...iv.linux.org.uk,
        mingo@...hat.com, peterz@...radead.org, hannes@...xchg.org,
        vbabka@...e.cz, mhocko@...e.com, vdavydov.dev@...il.com,
        minchan@...nel.org, riel@...hat.com, zhouxiyu@...wei.com,
        zhangshiming5@...wei.com, won.ho.park@...wei.com,
        tuxiaobing@...wei.com
Subject: Re: [PATCH] bdi flusher should not be throttled here when it fall
 into buddy slow path

On Tue, Oct 18, 2016 at 03:12:45PM +0800, zhouxianrong@...wei.com wrote:
> From: z00281421 <z00281421@...esmail.huawei.com>
> 
> bdi flusher may enter page alloc slow path due to writepage and kmalloc. 
> in that case the flusher as a direct reclaimer should not be throttled here
> because it can not to reclaim clean file pages or anaonymous pages
> for next moment; furthermore writeback rate of dirty pages would be
> slow down and other direct reclaimers and kswapd would be affected.
> bdi flusher should be iosceduled by get_request rather than here.
> 
> Signed-off-by: z00281421 <z00281421@...esmail.huawei.com>

What does this patch do that PF_LESS_THROTTLE is not doing already if
there is an underlying BDI?

There have been a few patches like this recently that look like they might
do something useful but are subtle. They really should be accompanied by
a test case and data showing they either fix a functional issue (machine
livelocking due to writeback not making progress) or a performance issue.

-- 
Mel Gorman
SUSE Labs