Message-ID: <487F0996.9010904@jp.fujitsu.com>
Date: Thu, 17 Jul 2008 17:57:58 +0900
From: Hidetoshi Seto <seto.hidetoshi@jp.fujitsu.com>
User-Agent: Thunderbird 2.0.0.14 (Windows/20080421)
MIME-Version: 1.0
To: Li Zefan <lizf@cn.fujitsu.com>
CC: Paul Jackson <pj@sgi.com>, LKML <linux-kernel@vger.kernel.org>,
       Paul Menage <menage@google.com>, Peter Zijlstra <peterz@infradead.org>,
       Andrew Morton <akpm@linux-foundation.org>,
       Lai Jiangshan <laijs@cn.fujitsu.com>
Subject: Re: [RFC] [PATCH] cpuset: fix wrong calculation of relax domain level
References: <487EFDD0.2060101@cn.fujitsu.com>
In-Reply-To: <487EFDD0.2060101@cn.fujitsu.com>
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 1290
Lines: 31

Li Zefan wrote:
> When multiple cpusets are overlapping in their 'cpus' and hence they
> form a single sched domain, the largest sched_relax_domain_level among
> those should be used. But when top_cpuset's sched_load_balance is
> set, its sched_relax_domain_level is used regardless other sub-cpusets'.
> 
> There are several proposals to solve this:
> 
> 1) Travel the cpuset hierarchy to find the largest relax_domain_level
>    in rebuild_sched_domains(). But cpuset avoids hierarchy travelling
>    when top_cpuset.sched_load_balance is set.
> 
> 2) Remember the largest relax_domain_level when we update a cpuset's
>    sched_load_balance, sched_relax_domain_level and cpus. This should
>    work, but seems a bit tricky and a bit ugly. (As this patch shows)
> 
> 3) Don't treat this as a bug, but document this behavior.

I think 1) is correct way.

There was a special short path for the top_cpuset's case, but now it is
disappeared. I think there are no need to treat the top_cpuset as VIP,
so 2) is excessive nurturing.

Thanks,
H.Seto
--
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/