Re: [Xenomai] Support of Beagleboard xm rev C on 3.14-ipipe

[Gilles Chanteperdrix <gilles.chanteperdrix@xenomai.org>]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 06/30/2014 02:01 PM, Arnaud Degroote wrote:
> On 30/Jun - 12:16, Gilles Chanteperdrix wrote:
>> -----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
>> 
>> On 06/30/2014 11:13 AM, Arnaud Degroote wrote:
>>> On 30/Jun - 10:59, Gilles Chanteperdrix wrote:
>>>> -----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
>>>> 
>>>> On 06/30/2014 10:26 AM, Arnaud Degroote wrote:
>>>>> On 27/Jun - 18:27, Gilles Chanteperdrix wrote:
>>>>>> -----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
>>>>>> 
>>>>>> On 06/27/2014 09:22 AM, Arnaud Degroote wrote:
>>>>>>> On 26/Jun - 19:27, Gilles Chanteperdrix wrote:
>>>>>>>> -----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
>>>>>>>> 
>>>>>>>> On 06/26/2014 01:19 PM, Arnaud Degroote wrote:
>>>>>>>>> On 26/Jun - 13:08, Gilles Chanteperdrix wrote:
>>>>>>>>>> switchtest is another story. Does switchtest
>>>>>>>>>> hang, or does the system completely freeze?
>>>>>>>>> 
>>>>>>>>> switchtest hangs, but the system is still
>>>>>>>>> responsive. I can ctrl-c it and restart it. On C-c,
>>>>>>>>> it prints "properly" a number of ctx switches done
>>>>>>>>> (which is far from other "normal" values).
>>>>>>>>> 
>>>>>>>> You mean it is as if switchtest had continued
>>>>>>>> switching contexts but had not printed anything? Do
>>>>>>>> you see counters incrementing in /proc/xenomai/stat
>>>>>>>> ?
>>>>>>> 
>>>>>>> I just meant that switchtest seems to respond properly
>>>>>>> to C-c, even if it does not progress. But counters from
>>>>>>>  /proc/xenomai/stat stops incrementing at some point.
>>>>>>> 
>>>>> 
>>>>> Related to my previous mail, I'm able to reduce threadspec
>>>>> to
>>>>> 
>>>>> rtk0 rtk0  rtup0
>>>>> 
>>>>> or
>>>>> 
>>>>> rtk0  rtup0 rtup0
>>>>> 
>>>>> Both demonstrates some quick hang. Other pattern may block
>>>>> too, I don't try the full combinatorial.
>>>>> 
>>>> Could you test rtk0 alone?
>>>> 
>>>> It would mean there is an issue with the context switch. Are
>>>> you running with the xenomai git?
>>> 
>>> It fails with 'rtk0 rtk0 rtk0' (really quickly). With 'rtk0
>>> rtk0' it seems to work.
>>> 
>>> I'm using xenomai-2.6 git 
>>> (4eb593df1eb89a47f1bc3423cea44c345d22ee3e).
>>> 
>> Ok, I just tested your configuration on my omap3 board, and
>> despite the config having a lot of stuff enabled that you
>> probably do not want, and may even cause issues (like memory
>> compaction), I did not observe any problem. I only had to enable
>> CONFIG_PINCTRL_SINGLE to have the serial console.
> 
> I can also confirm that the image (kernel + userland) works
> perfectly on my gumstix overo (I just changed uboot).
> 
> CPU: ARMv7 Processor [411fc083] revision 3 (ARMv7), cr=10c53c7d 
> Machine: Gumstix Overo OMAP3430/3530 ES3.1 (l2cache iva sgx neon
> isp )
> 
> 
> (except the two oops described previously in this tread).

Which two oops?

> 
> So, it may be related to this specific board ?
> 
> CPU: ARMv7 Processor [413fc082] revision 2 (ARMv7), cr=10c53c7d 
> Machine: OMAP3 Beagle Board OMAP3630 ES1.2 (l2cache iva sgx neon
> isp 192mhz_clk )

Have you checked that the exact same kernel configuration, only
without CONFIG_IPIPE and CONFIG_XENOMAI runs correctly under load?

- -- 
                                                                Gilles.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.12 (GNU/Linux)
Comment: Using GnuPG with Icedove - http://www.enigmail.net/

iD8DBQFTsVMfGpcgE6m/fboRAtcJAKCbXq6XWN5xLlzVevVgwY4IhwK1LwCfcqxE
XcdmKxxg4T1IbUwL7gVMCdU=
=jUeq
-----END PGP SIGNATURE-----