From mboxrd@z Thu Jan  1 00:00:00 1970
Message-ID: <51F2C9CF.6060206@xenomai.org>
Date: Fri, 26 Jul 2013 21:11:11 +0200
From: Gilles Chanteperdrix <gilles.chanteperdrix@xenomai.org>
MIME-Version: 1.0
References: <51B73421.3050506@siemens.com> <51EA7539.3000805@xenomai.org>
	<51F2A6AC.1060408@siemens.com>
In-Reply-To: <51F2A6AC.1060408@siemens.com>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit
Subject: Re: [Xenomai] Stable update for ipipe-3.8
List-Id: Discussions about the Xenomai project <xenomai.xenomai.org>
List-Unsubscribe: <http://www.xenomai.org/mailman/options/xenomai>,
	<mailto:xenomai-request@xenomai.org?subject=unsubscribe>
List-Archive: <http://www.xenomai.org/pipermail/xenomai>
List-Post: <mailto:xenomai@xenomai.org>
List-Help: <mailto:xenomai-request@xenomai.org?subject=help>
List-Subscribe: <http://www.xenomai.org/mailman/listinfo/xenomai>,
	<mailto:xenomai-request@xenomai.org?subject=subscribe>
To: Jan Kiszka <jan.kiszka@siemens.com>
Cc: Xenomai <xenomai@xenomai.org>

On 07/26/2013 06:41 PM, Jan Kiszka wrote:

> On 2013-07-20 13:32, Gilles Chanteperdrix wrote:
>> On 06/11/2013 04:28 PM, Jan Kiszka wrote:
>>
>>> Hi all,
>>>
>>> I've merged 3.8.13 into our stable 3.8 branch. This triggered some minor
>>> conflicts in generic and ARM code. Please see
>>>
>>> git://git.xenomai.org/ipipe-jki for-upstream/3.8
>>>
>>> for details. The result works fine for x86, and if it's also OK for the
>>> other arch, I'd propose this to be merged into the official 3.8 stable
>>> branch.
>>
>>
>> Hi Jan,
>>
>> I am not so sure 3.8.13. I am having random segfaults when running:
>>
>> xeno-regression-test -l "dohell -s 192.168.0.5 -m /mnt -l /ltp" -p 1000
>>
>> on geode and at91. I have first focused on at91 because the "unlocked
>> context switch" features is a usual suspect, but:
>> - the hardware pgd is in sync with the software pgd, which means that
>> the unlocked context switch did not mess up the hardware pgd;
>> -  going back to 3.5, I am unable to trigger the faults, despite the
>> fact that the mm switch code is the same for this type of machine.
>>
>> So, could you run this same test on your x86 machines to see if you have
>> any issue?
>
> I've seen it once during a test run today, but I'm lacking LTP here.


I forgot to add that yes, the issue is rare. It happens once in the 
whole run here, I happen to have noticed it because sometimes, the 
latency test is the victim. LTP exists as a debian package, if you are 
using debian.


-- 
                                                                 Gilles.