[FUSION-0.8.1]Bug with accessing RT_SEM from user space

[Philippe Gerum]

Kern Alexander wrote:
> Hello,
>  
> I observe a following behaviour on fusion-0.8.3 and fusion-0.8.1.
>  
> Userspace application forks a RT thread, which reads some datas, 
> syncronizing due a RT_SEM.
>  
> rtthread_fct() {
>     RT_SEM sem;
>  
>     rt_sem_bind(&sem);

0.8.3 and earlier releases have a nasty bug in the binding support code, 
that may cause rt_sem_bind() to return a spurious handle. You may want 
to give a try to the CVS version, just to verify this.

>     rt_sem_p(&sem); //USE IT AS synchronize point, this will be 
> permamently kicked from driver
>     //do some work
>     rt_sem_unbind(&sem);
>  
>     printf("YOU CAN SEE THIS LINE AS LAST, AFTER IT, APPLICATION STAYS\n")
>     printf("YOU CAN SIGABORT IT, BUT LINUX WILL BE SLOWLY AND SLOWLY, 
> AND WILL BE DIE!\n);
>     rt_task_delete(NULL);
> }
>  
> main() {
>     rt_tsk_spawn(...rttherad_fct);
>  
>     while(!done) {
>         printf("YOU CAN SEE THIS LINE!\n);
>         sleep(1);
>         printf("THIS CODELINE WILL BE NOT REACHED IF WE TOUCH RT_SEM, !\n);
>     }
> }
>  
> If I comment out lines with rt_sem_bind, rt_sem_p rt_sem_unbind, it 
> works like a charm!
>  
> Have anybody an idee, what is in my code wrong? Please help!
>  
> Kind Regards
>  
> Alexander Kern


-- 

Philippe.