Re: [Xenomai] __xn_sys_current_info failed: Operation not permitted

[Franz Engel <franz_lambert_engel@yahoo.de>]

Von: Philippe Gerum <rpm@xenomai.org>

An: Franz Engel <franz_lambert_engel@yahoo.de> 
>Gesendet: 15:03 Montag, 22.Juli 2013
>Betreff: Re: [Xenomai] __xn_sys_current_info failed: Operation not permitted
> 
>
>On 07/22/2013 02:10 PM, Franz Engel wrote:
>>
>>
>>     On 07/19/2013 12:57 PM, Franz Engel wrote:
>>
>>      >
>>      >    On 07/19/2013 11:49 AM, Franz Engel wrote:
>>      >      > Philippe wrote:
>>      >      >>> Franz Engel wrote:
>>      > >>> Hi,
>>      >      >> >
>>      >      >>> I get this error when I run a programm:
>>      >      >>> __xn_sys_current_info failed: Operation not permitted
>>      >      >> >
>>      > 
     >>> I know that the error comes from the assert_nrt_inner(void)
>>      >    function inside the assert_context.c file. But I have no idea what
>>      >    is the reason. The Operation not permitted is the return value
>>     from
>>      >    XENOMAI_SYSCALL1(__xn_sys_current_info, &info);. But I found no
>>      >    documentation about this function.
>>      >      >
>>      >      >
>>      >      >>This is reminiscent of a symptom observed for a bug fixed
>>     in the
>>      >    -forge
>>      >      >>tree, a while ago:
>>     
 >
>>      >>http://git.xenomai.org/?p=xenomai-forge.git;a=commit;h=014898add6a697e1a0d69f9e20b910ade74f1940
>>      >      >
>>      >      >>Do you happen to call fork() in your  application?
>>      >      >
>>      >      > Yes. I do? What I can do to solve the problem. Should I
>>     try to modify
>>      >      > the xenomai files as discrept in the git-link?
>>      >      >
>>      >
>>      >    I'm unsure, although this would be possible to backport
>>    
 portions of the
>>      >    logic, the implementation has now diverged significantly
>>     between the
>>      >    2.x
>>      >    and 3.x code bases.
>>      >
>>      >    Could you try the patch below? It's untested, but should work
>>     around
>>      >    the
>>      >    root issue in your case:
>>      >
>>      >    diff --git a/include/nucleus/compiler.h
>>     b/include/nucleus/compiler.h
>>      >    index 30f32c7..da4ea03 100644
>>      >    --- a/include/nucleus/compiler.h
>>      >    +++
 b/include/nucleus/compiler.h
>>      >    @@ -37,9 +37,11 @@
>>      >
>>      >        #ifdef __IN_XENO__
>>      >        #if __GNUC__ == 4 && __GNUC_MINOR__ == 6
>>      >    -#define __constructor__ __attribute__((constructor, noclone))
>>      >    +#define __constructor__ __attribute__((constructor(200),
>>     noclone))
>>      >    +#define __late_constructor__ __attribute__((constructor,
>>     noclone))
>>      >        #else /* Gcc not 4.6 */
>>      >    -#define __constructor__ __attribute__((constructor))
>>      >   +#define __constructor__
 __attribute__((constructor(200)))
>>      >    +#define __late_constructor__ __attribute__((constructor))
>>      >        #endif /* Gcc not 4.6 */
>>      >        #endif /* In xenomai */
>>      >
>>      >    diff --git a/src/skins/common/rt_print.c
>>     b/src/skins/common/rt_print.c
>>      >    index dfa8b33..e056cd6 100644
>>      >    --- a/src/skins/common/rt_print.c
>>      >    +++ b/src/skins/common/rt_print.c
>>      >    @@ -738,7 +738,7 @@ static void forked_child_init(void)
>>      >                spawn_printer_thread();
>>     
 >    }
>>      >
>>      >    -static __constructor__ void __rt_print_init(void)
>>      >    +static __late_constructor__ void __rt_print_init(void)
>>      >        {
>>      > const char *value_str;
>>      >            unsigned long long period;
>>      >
>>      >    --
>>      >    Philippe.
>>      >
>>      >    I tried the patch. But I think it is not the correct version of
>>      >    xenomai. In my compiler.h there is no line "#define
>>     __constructor__
>>      >    __attribute__((constructor,
 noclone))". I use Xenomai 2.6.2.1.
>>      >
>>
>>     With 2.6.2.1, you need the patch below applied, before applying the
>>     suggested work around:
>>     6e89cb2287378a7dcbba987949d0168460dabbe7
>>
>>     Bottom line is that all library constructor routines shall be tagged
>>     with the __constructor__ qualifier, except rt_print_init which
>>     should be
>>     __late_constructor__ instead. Otherwise the change makes no sense.
>>
>>     --
>>     Philippe.
>>
>>
>> O.k. I used the revision "6e89cb2287378a7dcbba987949d0168460dabbe7" and
>> I use the patch (the workaround). But the failure is still the same. Any
>> other idea?
>
>Do you mention --with-__thread in your configuration line? If so, please 
>retry without.
>
>-- 
>Philippe.
>
>
>
Do you 
mean the configuration for xenomai befor I install it? I don't use any 
parameter. Is there a way to trace such a failure. I get the error from a programm, that is not developed be my selfe.