Re: [Xenomai] Raspberry and Beaglebone patches

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

On 05/10/2013 01:53 PM, Stephan Kappertz wrote:

> Am 08.05.2013 um 20:27 schrieb Gilles Chanteperdrix:
> 
>> On 05/03/2013 10:41 PM, Stephan Kappertz wrote:
>>
>>>
>>> Am 03.05.2013 um 20:19 schrieb Gilles Chanteperdrix:
>>>
>>>> On 05/03/2013 04:28 PM, Stephan Kappertz wrote:
>>>>
>>>>>
>>>>> Am 02.05.2013 um 20:34 schrieb Gilles Chanteperdrix:
>>>>>
>>>>>> On 05/02/2013 02:19 PM, Stephan Kappertz wrote:
>>>>>>
>>>>>>>
>>>>>>> Am 16.04.2013 um 20:06 schrieb Gilles Chanteperdrix:
>>>>>>>
>>>>>>>> On 04/16/2013 12:26 PM, Stephan Kappertz wrote:
>>>>>>>>
>>>>>>>>> Am 15.04.2013 um 08:57 schrieb Gilles Chanteperdrix:
>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> Hi Stephan, Paul,
>>>>>>>>>>
>>>>>>>>>> we have a problem with the patches for Raspberry and 
>>>>>>>>>> BeagleBone: they are based on the 3.2.21 patch, which
>>>>>>>>>> we know is broken. So, we can not really ship the next
>>>>>>>>>> version of Xenomai with these patches.
>>>>>>>>>>
>>>>>>>>>> There are two ways we could keep patches for these 
>>>>>>>>>> machines: - you could send patches for 3.4 or 3.5, or
>>>>>>>>>> even 3.8 - we could backport the fixes to 3.2, but as
>>>>>>>>>> already discussed on this list, this would cost some
>>>>>>>>>> time to re-validate the 3.2 patch on all architectures.
>>>>>>>>>> However, if you do the backport, validate it on your
>>>>>>>>>> side, I could validate the result on my test boards,
>>>>>>>>>> and we would release a new version of the patch for 3.2
>>>>>>>>>> only for the ARM architecture.
>>>>>>>>>>
>>>>>>>>>> Regards.
>>>>>>>>>>
>>>>>>>>>> -- Gilles.
>>>>>>>>>>
>>>>>>>>>
>>>>>>>>> Hi Gilles,
>>>>>>>>>
>>>>>>>>> there's a 3.8.7 branch available for beagleBone at 
>>>>>>>>> https://github.com/beagleboard/kernel/tree/3.8.
>>>>>>>>>
>>>>>>>>> I can look into updating the beagleBone patch for that 
>>>>>>>>> repository. Where do I find the latest ipipe patch for
>>>>>>>>> 3.8 on ARM?
>>>>>>>>
>>>>>>>>
>>>>>>>> ipipe-gch repository, for-core-3.8 branch.
>>>>>>>>
>>>>>>>> -- Gilles.
>>>>>>>>
>>>>>>>
>>>>>>> Hi Gilles, do you have a script to create the ipipe patch
>>>>>>> file? Can't find any in the pipe-gch repository.
>>>>>>
>>>>>>
>>>>>> It is not yet in the repository, if you want the definitive
>>>>>> patch, please wait a bit, the patch should be released soon.
>>>>>>
>>>>>>
>>>>>> -- Gilles.
>>>>>>
>>>>>
>>>>> Hi Gilles,
>>>>>
>>>>> I have created a ipipe post patch for the beagleBone kernel
>>>>> v3.8.10 found here:
>>>>>
>>>>> https://github.com/RobertCNelson/linux-dev/tree/am33x-v3.8,
>>>>> branch am33x-v3.8
>>>>>
>>>>> Your pipe-gch repository is based on 3.8 while the upper
>>>>> repository is at 3.8.10. Thus I had to apply a few minor changes
>>>>> to the patch file created from your git. Are you going to rebase
>>>>> the ipipe patch to 3.8.10 or higher before releasing it?
>>>>
>>>>
>>>> I do not know, Philippe is going to make the patch, he will decide
>>>> on which version he rebases.
>>>>
>>>>> If so, the beagleBone patch gets reduced to the post patch
>>>>> below:
>>>>
>>>>
>>>> Ok, several remarks: - can not we put that code in the mainline
>>>> kernel (meaning, does soc_is_am33xx() exist in the mainline kernel,
>>>> if it does not exist, can not we add it?) - the coding style for
>>>> the tsc declaration is bad, braces are not put on separate lines in
>>>> the Linux kernel coding style; - since the only difference between
>>>> am33xx and other omaps is OMAP_TIMER_CAPTURE_REG vs
>>>> OMAP_TIMER_COUNTER_REG, why not put this in a local variable and
>>>> use it twice?
>>>>
>>>>
>>>> -- Gilles.
>>>>
>>>
>>> Yes, soc_is_am33xx() is indeed defined in the mainline kernel
>>> starting from 3.6. So the beagleBone patch can safely be put into the
>>> mainline ipipe sources.
>>>
>>> I'll fix the coding style and send a new patch. Actually using
>>> soc_is_am33xx() for the TSC register location decision was an
>>> unnecessary constraint. The OMAP timer struct has a revision field
>>> that is 2 for the v2 timer IP. In rev 2, all functional registers are
>>> offset by OMAP_TIMER_V2_FUNC_OFFSET.
>>
>>
>> Hi,
>>
>> any news about this patch?
>>
>> Regards.
>>
>> -- 
>>                                                                Gilles.
>>
> 
> Sorry for the delay, here it is. Tested against pipe-gch for-core-3.8. Xenomai is successfully starting on the beagleBone:


What version of Linux are you using, is this 3.8.0, or another version
as you told me in a previous mail?

-- 
                                                                Gilles.