linux-kernel - RE: [PATCH] mm: vmscan: avoid a unnecessary reschedule in shrink

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: <9a2b772b13f84bdd9517b17d8d72aa89@h3c.com>
Date:   Fri, 16 Oct 2020 12:48:23 +0000
From:   Tianxianting <tian.xianting@....com>
To:     Michal Hocko <mhocko@...e.com>
CC:     "akpm@...ux-foundation.org" <akpm@...ux-foundation.org>,
        "linux-mm@...ck.org" <linux-mm@...ck.org>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>
Subject: RE: [PATCH] mm: vmscan: avoid a unnecessary reschedule in
 shrink_slab()

Thanks, my understanding is,
In shrink_slab(), do_shrink_slab() will do the real reclaim work, which will occupy current cpu and consume more cpu time, so we need to trigger a reschedule after reclaim.
But if it jumps to 'out' label, that means we don't do the reclaim work at this time, it won't cause other thread getting starvation, so we don't need to call cond_resched() in this case.
Is it right?

-----Original Message-----
From: Michal Hocko [mailto:mhocko@...e.com] 
Sent: Friday, October 16, 2020 8:08 PM
To: tianxianting (RD) <tian.xianting@....com>
Cc: akpm@...ux-foundation.org; linux-mm@...ck.org; linux-kernel@...r.kernel.org
Subject: Re: [PATCH] mm: vmscan: avoid a unnecessary reschedule in shrink_slab()

On Fri 16-10-20 11:39:52, Xianting Tian wrote:
> In shrink_slab(), it directly goes to 'out' label only when it can't 
> get the lock of shrinker_rwsew. In this case, it doesn't do the real 
> work of shrinking slab, so we don't need trigger a reschedule by 
> cond_resched().

Your changelog doesn't explain why this is not needed or undesirable. Do you see any actual problem?

The point of this code is to provide a deterministic scheduling point regardless of the shrinker_rwsew.

> 
> Signed-off-by: Xianting Tian <tian.xianting@....com>
> ---
>  mm/vmscan.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/mm/vmscan.c b/mm/vmscan.c index 466fc3144..676e97b28 
> 100644
> --- a/mm/vmscan.c
> +++ b/mm/vmscan.c
> @@ -687,8 +687,9 @@ static unsigned long shrink_slab(gfp_t gfp_mask, int nid,
>  	}
>  
>  	up_read(&shrinker_rwsem);
> -out:
> +
>  	cond_resched();
> +out:
>  	return freed;
>  }
>  
> --
> 2.17.1
> 

-- 
Michal Hocko
SUSE Labs