Message-ID: <55550BC5.4060906@ezchip.com>
Date: Thu, 14 May 2015 16:55:33 -0400
From: Chris Metcalf <cmetcalf@ezchip.com>
User-Agent: Mozilla/5.0 (X11; Linux i686 on x86_64; rv:31.0) Gecko/20100101 Thunderbird/31.6.0
MIME-Version: 1.0
To: <paulmck@linux.vnet.ibm.com>, Peter Zijlstra <peterz@infradead.org>
CC: Gilad Ben Yossef <giladb@ezchip.com>, Steven Rostedt <rostedt@goodmis.org>,
        Ingo Molnar <mingo@redhat.com>,
        Andrew Morton <akpm@linux-foundation.org>,
        Rik van Riel <riel@redhat.com>, Tejun Heo <tj@kernel.org>,
        Frederic Weisbecker <fweisbec@gmail.com>,
        Thomas Gleixner <tglx@linutronix.de>, Christoph Lameter <cl@linux.com>,
        <linux-kernel@vger.kernel.org>
Subject: Re: [PATCH 2/6] nohz: dataplane: allow tick to be fully disabled
 for dataplane
References: <1431107927-13998-1-git-send-email-cmetcalf@ezchip.com> <1431107927-13998-3-git-send-email-cmetcalf@ezchip.com> <20150512092607.GE21418@twins.programming.kicks-ass.net> <20150512131250.GJ6776@linux.vnet.ibm.com>
In-Reply-To: <20150512131250.GJ6776@linux.vnet.ibm.com>
Content-Type: text/plain; charset="windows-1252"; format=flowed
Content-Transfer-Encoding: 7bit
X-MS-Exchange-CrossTenant-OriginalArrivalTime: 14 May 2015 20:55:47.7249 (UTC)
X-MS-Exchange-CrossTenant-FromEntityHeader: Hosted
X-MS-Exchange-Transport-CrossTenantHeadersStamped: AM2PR02MB0772
Sender: linux-kernel-owner@vger.kernel.org
List-ID: <linux-kernel.vger.kernel.org>
X-Mailing-List: linux-kernel@vger.kernel.org
Content-Length: 2247
Lines: 47

On 05/12/2015 09:12 AM, Paul E. McKenney wrote:
> On Tue, May 12, 2015 at 11:26:07AM +0200, Peter Zijlstra wrote:
>> On Fri, May 08, 2015 at 01:58:43PM -0400, Chris Metcalf wrote:
>>> While the current fallback to 1-second tick is still helpful for
>>> maintaining completely correct kernel semantics, processes using
>>> prctl(PR_SET_DATAPLANE) semantics place a higher priority on running
>>> completely tickless, so don't bound the time_delta for such processes.
>>>
>>> This was previously discussed in
>>>
>>> https://lkml.org/lkml/2014/10/31/364
>>>
>>> and Thomas Gleixner observed that vruntime, load balancing data,
>>> load accounting, and other things might be impacted.  Frederic
>>> Weisbecker similarly observed that allowing the tick to be indefinitely
>>> deferred just meant that no one would ever fix the underlying bugs.
>>> However it's at least true that the mode proposed in this patch can
>>> only be enabled on an isolcpus core, which may limit how important
>>> it is to maintain scheduler data correctly, for example.
>> So how is making this available going to help people fix the actual
>> problem?
> It will at least provide an environment where adding more of this
> problem might get punished.  This would be an improvement over what
> we have today, namely that the 1HZ fallback timer silently forgives
> adding more problems of this sort.

So I guess the obvious question to ask is whether there is a mode
that can be dynamically enabled (/proc/sys/kernel/nohz_experimental
or whatever) where we allow turning off this tick - perhaps to make
it more likely tick-dependent code isn't added to the kernel as Paul
suggests, or perhaps to enable applications that want to avoid the
tick conservativeness and are willing to do sufficient QA that they
are comfortable exploring possible issues with the 1Hz tick being
disabled?

Paul, PeterZ, any thoughts on something along these lines?
Or another suggestion?

-- 
Chris Metcalf, EZChip Semiconductor
http://www.ezchip.com

--
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/