Experimental Physics and
Industrial Control System

Marty Kraimer <[email protected]> · Tue, 10 Jun 1997 11:00:48 -0500

Ralph Lange wrote:
> 
> Dear Gurus,
> 
>         since Friday we're observing an unpleasant behaviour on one of our
> IOCs:
>         Occasionally (approx. every 20 to 30 minutes at normal load) the
> scanOnce task gets suspended due to an Access Fault. After a couple of
> seconds, the dbCaLink task follows after writing "rngBufPut overflow in
> scanOnce" to the log. After that there is no more ca based record
> processing, i.e. no more ca links and operator access.

Request are made to scanOnce by putting a request into a ring buffer.
thus when scanOnce crashes the ring buffer ultimately fills up. This is
what
you are seeing.

Note that NO epics tasks are automatically restarted when they crash.
Sounds nice but can cause more problems than it solves.

>         So - who the hell is calling scanOnce() with a junk record pointer
> (this is the way we think dbScanLock gets called)?

Good question. Answer should solve problem!!

>         We inserted a patch in scanOnce to check the precord argument for
> validity and start a tt() to see which task is the bad guy here. Another
> advantage of that patch is that we won't process the record and therefore
> extend the IOC's uptime.

So what did you find?

If you get a stack trace of crashed scanOnce task it should help.

Marty

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System