linux-kernel - Re: [PATCH 4/4] sched/numa: Do not move imbalanced load purely on the basis of an idle CPU

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]

Date:   Fri, 7 Sep 2018 15:28:46 +0100
From:   Mel Gorman <mgorman@...hsingularity.net>
To:     Srikar Dronamraju <srikar@...ux.vnet.ibm.com>
Cc:     Peter Zijlstra <peterz@...radead.org>,
        Ingo Molnar <mingo@...nel.org>,
        Rik van Riel <riel@...riel.com>,
        LKML <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH 4/4] sched/numa: Do not move imbalanced load purely on
 the basis of an idle CPU

On Fri, Sep 07, 2018 at 07:12:01PM +0530, Srikar Dronamraju wrote:
> > Yeah, I was afraid it would.. Srikar, can you also evaluate, I suspect
> > we'll have to pick one of these two patches.
> 
> I can surely run some benchmarks between the two patches.
> However comparing  Mel's patch with
> http://lkml.kernel.org/r/1533276841-16341-4-git-send-email-srikar@linux.vnet.ibm.com
> 
> Mel's patch
> 
>   	if (!cur) {
>  -		if (maymove || imp > env->best_imp)
>  +		if (maymove)
>   			goto assign;
>   		else
> http://lkml.kernel.org/r/1533276841-16341-4-git-send-email-srikar@linux.vnet.ibm.com
> 
> 
>  	if (!cur) {
> -		if (maymove || imp > env->best_imp)
> +		if (maymove && moveimp >= env->best_imp)
>  			goto assign;
>  		else
> 
> In Mel's fix, if we already found a candidate task to swap and then encounter a
> idle cpu, we are going ahead and overwriting the swap candidate. There is
> always a chance that swap candidate is a better fit than moving to idle cpu.
> 

There is a chance but to find out, the task has to be dequeued and requeued
on a maybe. An idle CPU is less disruptive and the only task affected is
migrating to the preferred node where, based on previous fault behaviour,
should have better locality. It's also a simplier patch but I'm going to
be biased towards my own patch, the tests will decide one way or the other.

-- 
Mel Gorman
SUSE Labs