RTEMS Q returns RTEMS_SUCCESSFUL but with length of message zero

[Joel Sherrill]

John Bebbington wrote:
> 
> Hi,
> 
> I have been using RTEMS for some time now with very good results.
> Unfortunately I still have an unexplained event which I currently have a
> work-around for in the application but would like to ask if perhaps the
> behaviour is correct or not.

You are always beating the message queue up. :)

> When using rtems_message_queue_receive(..) with no blocking (polling) I get
> a return code RTEMS_SUCCESSFUL but when I check the length of the message it
> is zero!

What version of RTEMS are you using?  Just in case something has been
reported since this.

What specific calls are you making at the task/receive level and
ISR/send
level?  

What CPU are you on?  There was a recent patch against the m68k to
address a
VERY m68k specific problem in honoring RTEMS protocol in recording ISR
level.  

My gut feeling is to do a source examination for potential windows.

> The example below shows the call and the additional plaus test I am doing as
> a work-around.
> 
> Additional Information:
> ========================
> 
> o The system is being interrupted by about 5 levels of interrupt handlers
> which all are attached to RTEMS via the
>   rtems_interrupt_catch directive.
> o When The CPU duty is approx 12% the error does not occur, but with
> additional transmits and a duty of 49% the
>   errors start occurring.
> 
> The question is, is the call to rtems_message_queue_receive() "managed" to
> the extend that the RTEMS_SUCESSFUL can only be returned when there is a
> message available.
> 
> -----start of code snippit ---------------
> 
>         rtems_status_code       status  ;
>         unsigned32                      size    =       0;
>         bool                            newmsg;
> 
>         // Check for new messages on the input Q but do not WAIT!!!
>         status = (
>                         QueueID[q_BIT_HISTORY_Q],
>                         rec,
>                         &size,
>                         RTEMS_NO_WAIT,
>                         0
>                         );
> 
>         switch (status){
>                 case RTEMS_SUCCESSFUL:
>                         newmsg = true;
>                         break;
>                 case RTEMS_UNSATISFIED:
>                         newmsg = false;
>                         break;
>                 default:
>                         newmsg = false;
>                         ON_RTEMS_ERROR_LOG2( status, E_RTEMS_Q_READ, 0x0001 );
>                         break;
>         };
> 
>         if ( newmsg && (size == 0) ){
> 
>                 // Extra check for plausibility.
>                 // =============================
>                 // We sometimes get a message with length = 0
>                 // but with status RTEMS_SUCCESSFUL, this is not logical.
>                 // In this case we shall ignore the zeroed message and log the error.
> 
>                 newmsg = false;
>                 LOG2( E_RTEMS_Q_RETURNS_OK_BUT_LEN_ZERO,0x0001);
>         }
> 
> -----end of code snippit ---------------
> 
> John Bebbington
> Freelance Software Engineer
> Black Forest
> Germany
> web-site : www.rtss.de

-- 
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