Re: [Xenomai] Hilscher driver for cifX boards

[Jan Kiszka <jan.kiszka@siemens.com>]

On 2013-03-15 10:09, Jerome Poncin wrote:
> Le 13/03/2013 12:08, Jerome Poncin a écrit :
>> Le 12/03/2013 20:38, Gilles Chanteperdrix a écrit :
>>> On 03/12/2013 04:27 PM, Jerome Poncin wrote:
>>>
>>>> Le 12/03/2013 13:21, Gilles Chanteperdrix a écrit :
>>>>> On 03/12/2013 10:10 AM, Jerome Poncin wrote:
>>>>>
>>>>>> Le 08/03/2013 13:22, Gilles Chanteperdrix a écrit :
>>>>>>> On 03/08/2013 11:17 AM, Jerome Poncin wrote:
>>>>>>>
>>>>>>>> Le 07/03/2013 16:33, Jerome Poncin a écrit :
>>>>>>>>> Le 06/03/2013 22:05, Gilles Chanteperdrix a écrit :
>>>>>>>>>> On 03/06/2013 04:28 PM, Jan Kiszka wrote:
>>>>>>>>>>
>>>>>>>>>>> Also here: Do not misuse write(), define an IOCTL that contains
>>>>>>>>>>> something like "MMAP" and another one with MUNMAP in its name
>>>>>>>>>>> - that's
>>>>>>>>>>> the purpose of this service, no?
>>>>>>>>>> A lot of drivers probably have to reinvent a way to pass all the
>>>>>>>>>> parameters for an mmap, why not adding an "mmap" method to
>>>>>>>>>> RTDM drivers?
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>> Hello,
>>>>>>>>>
>>>>>>>>> I  re factored my code with all your remarks (I think and I hope).
>>>>>>>>>
>>>>>>>>> Could you say me if it's correct now ?
>>>>>>>>>
>>>>>>>>> Jan, you're right, It's possible to use Xenomai with UIO
>>>>>>>>> driver, but I
>>>>>>>>> would like to use more real time possibilities of Xenomai in
>>>>>>>>> future
>>>>>>>>> (like IRQ).
>>>>>>>>> My first goal is to have a code compliant.
>>>>>>>>> I added IRQ, and my next step will be to understand why the kernel
>>>>>>>>> crash after IRQ registration.
>>>>>>>>>
>>>>>>>>> Thank you for your help,
>>>>>>>>>
>>>>>>>>> Jerome
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>> Hello,
>>>>>>>>
>>>>>>>> I have good news because, I arrived to validated driver with IRQ
>>>>>>>> support. I changed position of cifx on my motherboard in my PC
>>>>>>>> and the
>>>>>>>> IRQ affected is not shared ;-) !
>>>>>>>>
>>>>>>>> I have now a question about IRQ. What is the good method to
>>>>>>>> signal to
>>>>>>>> user-land that there is an IRQ. Is there an example ?
>>>>>>> Almost every "read" call is an example... ;-)
>>>>>>>
>>>>>> Hello,
>>>>>>
>>>>>> I'm finishing my driver with IRQ. I saw that there is two mode, Real
>>>>>> time and Non real time for RTDM driver function.
>>>>>> My driver should run in the two mode. When I compile my User Space
>>>>>> Shared library that call Xenomai Kernel driver,
>>>>>> I saw that library call every time Non real time function and no real
>>>>>> time function...
>>>>>> I use a shared library with posix skin with makefile in attached
>>>>>> file. I
>>>>>> read explanation for static library with posix skin :
>>>>>>
>>>>>> http://www.xenomai.org/index.php/Porting_POSIX_applications_to_Xenomai
>>>>>>
>>>>>>
>>>>>> What must I do with shared library for using rt mode ?
>>>>> Use the LDLIBS variable when linking the library?
>>>>>
>>>> Gilles,
>>>>
>>>> you're right ;-), I didn't use LDLIBS for linking the library,
>>>> therefore
>>>> I corrected this.
>>>>
>>>> Now I have this as link line for my library :
>>>>
>>>> gcc -shared -Xlinker -rpath -Xlinker /usr/xenomai/lib
>>>> -I/usr/xenomai/include/include/xenomai
>>>> -I/usr/xenomai/include/include/xenomai/compat
>>>> -I/usr/xenomai/include/include/xenomai/posix -I./ -I.//Toolkit
>>>> -Wl,@/usr/xenomai/lib/posix.wrappers -L/usr/xenomai/lib -lpthread_rt
>>>> -lxenomai -lpthread -lrt -o libcifx.so .//cifXDownload.o
>>>> .//cifXEndianess.o .//cifXFunctions.o .//cifXHWFunctions.o
>>>> .//cifXInit.o
>>>> .//cifXInterrupt.o .//md5.o   .//cifx_xenomai.o .//OS_xenomai.o
>>>> .//USER_xenomai.o
>>>>
>>>> and this for the application :
>>>>
>>>> gcc -DCONSUMER -I/usr/xenomai/include -D_GNU_SOURCE -D_REENTRANT
>>>> -D__XENO__ -I/usr/xenomai/include/posix -I../libcifx
>>>> -I../libcifx/Toolkit -O2   -lcifx -lpciaccess -lrtdm -Xlinker -rpath
>>>> -Xlinker /usr/xenomai/lib  cifx_xenomai_sample.c
>>>> -Wl,@/usr/xenomai/lib/posix.wrappers -L/usr/xenomai/lib -lpthread_rt
>>>> -lxenomai -lpthread -lrt -lcifx -lpciaccess -lrtdm -o
>>>> cifx_xenomai_sample
>>>>
>>>> with
>>>>
>>>> mlockall(MCL_CURRENT | MCL_FUTURE);
>>>>
>>>>
>>>> first line in main
>>>>
>>>> I have a new log :
>>>>
>>>> [ 4700.073616] cifx 0000:04:01.0: PCI INT A -> GSI 16 (level, low)
>>>> -> IRQ 16
>>>> [ 4700.073643] cifx 0000:04:01.0: DPM at FEBF0000
>>>> [ 4700.073654] cifx 0000:04:01.0: registered CifX card
>>>> [ 4734.146888] NRT mode
>>>> [ 4751.104009] rt
>>>> [ 4751.388002] Xenomai: Posix: destroying mutex f8419600.
>>>>
>>>> But as you see It seems I'm always in Non Real Time mode...
>>> I do not see anything, as I do not see the sources for either your
>>> driver or application.
>>>
>>>> Do you have an idea about my problem ?
>>>>
>>>> The log "rt" and " Xenomai: Posix: destroying mutex f8419600." is
>>>> normal ?
>>> It depends if you consider normal leaving an application with a mutex
>>> created but not destroyed.
>> Hello Gilles,
>>
>> First, I checked mutex Creation and Destroy and all is correct all
>> created are destroyed... I will continue to investigate, it's very
>> strange !
>> Secondly, I read that if a program use file or stdio library "causes
>> switches to secondary mode".
>> In my shared library, I have read file and fprintf at start
>> (initialization), therefore I think that's why the library run in NRT
>> mode.
>> But is possible to switch in RT mode after, or must I do two thread ?
>>
>> thank you for your help,
>>
>> Jerome
> Hello,
> 
> I found why I have the log :
> 
> Xenomai: Posix: destroying mutex f8419600.
> 
> My application seems stop by the watchdog :
> 
> Temps UCT limite expiré (core dumped)
> 
> I tried to disable Xenomai watchdog without success, and I try to
> increase "ulimit" without success too.
> 
> How to disable CPU time watchdog for real time mode ?

The problem is not the watchdog but some bug in your application or
driver. It causes a real-time task to run for multiple seconds without
sleeping. That is what the watchdog triggers. Check what is causing this
endless loop and resolve it.

Jan

-- 
Siemens AG, Corporate Technology, CT RTC ITP SDP-DE
Corporate Competence Center Embedded Linux