Re: [PATCH] sched: Fix numabalancing to work with isolated cpus

From: Michal Hocko <mhocko@kernel.org>
To: Peter Zijlstra <peterz@infradead.org>
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>,
	Ingo Molnar <mingo@kernel.org>,
	LKML <linux-kernel@vger.kernel.org>,
	Mel Gorman <mgorman@techsingularity.net>,
	Rik van Riel <riel@redhat.com>
Subject: Re: [PATCH] sched: Fix numabalancing to work with isolated cpus
Date: Thu, 6 Apr 2017 15:44:07 +0200	[thread overview]
Message-ID: <20170406134405.GN5497@dhcp22.suse.cz> (raw)
In-Reply-To: <20170406104757.kemq6is6nh6f26fx@hirez.programming.kicks-ass.net>

On Thu 06-04-17 12:47:57, Peter Zijlstra wrote:
> On Thu, Apr 06, 2017 at 12:42:04PM +0200, Michal Hocko wrote:
> 
> > Is this something dictated by usecases which rely on isolcpus or rather
> > nobody bothered to implement one scheduling domain?
> 
> Its from the original use-case I suspect. It was done very much on
> purpose.

OK, fair enough.

> If you want bigger partitions use cpusets. Or rather, like I've been
> saying for many years, use cpusets for everything and kill isolcpus.

I thought that the only point of isolcpus is to provide an exclusive set
of cpus that only selected userspace will ever run. Mostly to schield it
of from the random system activity. But maybe this is just my
misinterpretation of what this feature offers.

Anway I am shifting this off-topic so I will not interfere more. Thanks
for the clarification.
-- 
Michal Hocko
SUSE Labs