RE: Access Fault

[johill@lanl.gov (Jeff Hill)]

>      I'm running version 3.12.2 on a 167.  The program is relatively
> simple; it just reads out a stream of data from some Allen-Bradley
> modules.  It had been running stably a few months ago.  I just turned the
> processor back on yesterday, and within five minutes, I got the following
> Access Fault message:
>
> Access Fault
> Program Counter: 0x0008c91e
> Status Register: 0x3009
> Access Address : 0x01010101
> Special Status : 0x0505
> Task: 0x78a990 "tNetTask"
>
> after which, all channel access is lost.  Every time I reboot, I
> eventually get this message, sometimes with "(CA event): panic: mget"
> messages beforehand, sometimes not.  Can anyone help me with this?
> Thanks.
>

For what its worth, with little to start from, here is a wild and completely
speculative guess:

The access address of the exception (0x01010101) looks like a repetitive
pattern that could have been caused by an errant program with an
unterminated
write loop.

If the net task receives an unhandled access fault exception then this will
certainly shut down the WRS IP kernel. I would use "tt (<tNetTask task id>)"
to obtain a stack trace and hopefully determine what WRS code is involved.
The panic message generally indicates that some BSD derived code is having
troubles. In this case the message is occurring from the CA event task
probably because it is calling routines in the BSD socket library which are
printing the panic message. A wild guess is that "mget()" may be the
internal
function in the BSD IP kernel that allocates a fixed sized networking buffer
(an "mbuf").

Jeff