Executive shutdown

[Joel Sherrill <joel@OARcorp.com>]

brian wrote:
> Joel,
> 
> Thank you for your reply,
> 
> As part of my studies I am looking at the reuse of an application that was
> crafted directly to a processor and looking at how it could be moved to a
> more modern platform. Because of the close coupling possible it usd a 2ms
> interrupt to drive it.

With anything but POSIX signals, 2 milliseconds wouldn't be a big deal.
A more proper RTEMS solution would be to switch to Classic API Events
and set up a server task to receive them.  Much lower overhead.

> I have already found that a different processor/OS combintaion couldnot
> cope, the singals generated from the interrupt were being missed by the
> application because it could not keep up.
> 
> Is there any way to put a trace in to check for stack overflow? and how easy
> is it to increase the stack ?

There is a stack checker in cpukit/libmisc/stackchk.  Define
STACK_CHECKER_ON before including confdefs.h.  The checks at
each context switch.  You might have to insert a force check
at critical points.

> Thank You
> 
> ----- Original Message ----- 
> From: "Joel Sherrill <joel at OARcorp.com>" <joel.sherrill at OARcorp.com>
> To: "Stevens Brian" <BPSTEVENS at qinetiq.com>
> Cc: "'Angelo Fraietta'" <afraiett at bigpond.net.au>; <rtems-users at rtems.com>
> Sent: Friday, July 02, 2004 9:03 PM
> Subject: Re: Executive shutdown
> 
> 
> 
>>Stevens Brian wrote:
>>
>>>I have added a counter and it indicates that the task receives approx
> 
> 280
> 
>>>signals before failing.
>>
>>Is this a clean exit/shutdown, fatal error, or just a random fault?
>>
>>FWIW the POSIX signals were designed to be safe to be sent from
>>interrupts but their complexity has always made me suspicious of
>>doing so.  At your signal rate, you could be overflowing a task
>>stack.  Each signal is going to push deeper and deeper onto the
>>interrupted thread's stack.
>>
>>--joel
>>
>>
>>>-----Original Message-----
>>>From: Angelo Fraietta [mailto:afraiett at bigpond.net.au]
>>>Sent: 02 July 2004 09:39
>>>To: Stevens Brian
>>>Cc: 'rtems-users at rtems.com'
>>>Subject: Re: Executive shutdown
>>>
>>>
>>>
>>>Stevens Brian wrote:
>>>
>>>
>>>
>>>>RTEMS version 4.6
>>>>Target PC386
>>>>
>>>>I have written a device driver to use the rtc to generate timer
> 
> interrupts
> 
>>>>independent of the clock used to drive rtems. I have set  the additional
>>>>timer to produce interupts at 122micro second intervals, these are then
>>>>counted and when the required time interval has elapsed a POSIX signal
> 
> is
> 
>>>>sent to a thread.
>>>>
>>>>If I set the signal to be sent every  2 miliseconds the system
> 
> eventually
> 
>>>>crashes with an executive shutdown. However, if set to 20ms it works
> 
> fine.
> 
>>>>
>>>>
>>>After how long does it crash? Do you get a message saying that it shust
> 
> down
> 
>>
>>-- 
>>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