pthread's mutex question

[Joel Sherrill]

John Ford wrote:
> 
> Pattara Kiatisevi writes:
>  > On Mon, 29 Apr 2002, Joel Sherrill wrote:
>  >
>  > >
>  > >
>  > > Pattara Kiatisevi wrote:
>  > > >
>  > > > Hi again,
>  > > >
>  > > > I have played with pthread's mutex with this small code:
>  > > >
>  > > >   pthread_mutex_t a,b;
>  > > >
>  > > >   printf("Initialize mutexes...\n");
>  > > >   pthread_mutex_init(&a, NULL);
>  > > >   pthread_mutex_init(&b, NULL);
>  > > >
>  > > >   printf("Trying to lock mutex a..\n");
>  > > >   pthread_mutex_lock(&a);
>  > > >   printf("Got mutex a..\n");
>  > > >
>  > > >   printf("Trying to lock mutex a..\n");
>  > > >   pthread_mutex_lock(&a);       <----------should block here
>  > > >   printf("Got mutex a..\n");
>  > > >
>  > > >   fprintf(stderr,"Done.\n");
>  > > >
>  > > > On Linux, the program will block forever at the second attempt to lock
>  > > > mutex a but on RTEMS it doesn't block. Instead it returns with value 45.
>  > > > Is there any chance to have it blocked here in RTEMS?
>  > >
>  > > That's EDEADLOCK which probably means you are attempting to acquire
>  > > a mutex you already own.
>  >
>  > Yeah and normally the second lock should block? Or is there any adjustable
>  > parameter?
> 
> This is an undefined result using POSIX threads.  The second call can
> either return EDEADLOCK, or it can simply honor your request and
> deadlock the task.  It is implementation dependent, meaning that you
> should never rely on it working one way or the other.  Obviously RTEMS
> and Linux implement it differently.  I much prefer the RTEMS way!
> Silent deadlock is not my idea of fun...

I agree and it looks like the Open Group UNIX98 standard has
fine-tuned the application options a bit.  

RTEMS has a mutex attribute called "recursive" to set the desired
behavior but the general RTEMS philosophy is to not let you trip
over your own feet. :)
 
> John
> 
>  >
>  > > Also ... What target and BSP are you using?  If accidentally
>  > > using unix/posix that is an impact since RTEMS POSIX API
>  > > can't be used there.
>  >
>  > sparc-rtems/leon2.
>  >
>  > Thank you,
>  > Pattara
>  >
>  > >
>  > > > Regards,
>  > > > Pattara
>  > > >
>  > > > --
>  > > > Please avoid sending me Word or PowerPoint attachments.
>  > > > See http://www.fsf.org/philosophy/no-word-attachments.html
>  > > > ----------------------------------------------------------------------
>  > > > Ott Pattara Kiatisevi                              T L W G
>  > > > M.Sc. INFOTECH Student, Stuttgart, Germany      http://linux.thai.net/
>  > > > ----------------------------------------------------------------------
>  > >
>  > > --
>  > > Joel Sherrill, Ph.D.             Director of Research & Development
>  > > joel at OARcorp.com                 On-Line Applications Research
>  > > Ask me about RTEMS: a free RTOS  Huntsville AL 35805
>  > >    Support Available             (256) 722-9985
>  > >
>  > >
>  >
>  >

-- 
Joel Sherrill, Ph.D.             Director of Research & Development
joel at OARcorp.com                 On-Line Applications Research
Ask me about RTEMS: a free RTOS  Huntsville AL 35805
   Support Available             (256) 722-9985