RTEMS PowerPC progress.

[Nick Thomas]

OK, so I think I have come to the end of my RTEMS on PowerPC stability
issues.
My last email said that I would suspend the timer_server task at the top of
rtems_timer_server_fire_after, and re-enable it at the end.

This is still working OK. It has been one week since the last crash.

And, although this surely isn't the best solution, and I don't really
understand why this seems to fix it, I am going to move forward to
production with this modification in place.
1 week uptime is good enough for this application, because a crash won't
hurt anyone.


Thanks to everyone who helped get me this far.


Regards

Nick


> -----Original Message-----
> From: rtems-users-bounces at rtems.org [mailto:rtems-users-
> bounces at rtems.org] On Behalf Of Nick Thomas
> Sent: 04 June 2010 07:33
> To: rtems-users at rtems.org
> Subject: RE: RTEMS PowerPC progress.
> 
> >
> > OK, I have back-ported the timer server code from 4.9.4 in my 4.7.3
> > source
> > tree.
> > But, it still crashes.
> >
> > I have noticed that it behaves itself for longer if I make each call
> to
> > rtems_timer_server_fire_after preceeded by rtems_interrupt_disable(),
> > and
> > make the calling task RTEMS_NO_PREEMPT.
> > And putting the settings back immediately after the call.
> >
> > This has stopped interrupts getting in during the period after
> > _ISR_Enable,
> > which was causing the _Thread_Dispatch_disable_level to go negative.
> >
> > But, I suspect that the Timer_Server_body task itself is running
> during
> > this
> > period, and causing the _Thread_Dispatch_disable_level to go to zero,
> > and
> > the next call to _Thread_Enable_dispatch() in
> > rtems_timer_server_fire_after
> > is making it go negative.
> >
> 
> 
> OK, so I have made another test modification.
> At the top of rtems_timer_server_fire_after, I suspend the timer_server
> task, and resume it just before it returns.
> This prevents the task from running during that functions execution.
> 
> I know this doesn't sound like the right thing to do.
> 
> But, I now have two units on test overnight, and this morning both are
> running OK. So, this change isn't harmful - even if it isn't very
> elegant.
> 
> At this stage I am going to go with what works.
> 
> 
> Regards
> 
> Nick
> 
> 
> _______________________________________________
> rtems-users mailing list
> rtems-users at rtems.org
> http://www.rtems.org/mailman/listinfo/rtems-users
> No virus found in this incoming message.
> Checked by AVG - www.avg.com
> Version: 9.0.829 / Virus Database: 271.1.1/2914 - Release Date:
> 06/02/10 19:25:00