RE: ask for help about ca

["Jeff Hill" <johill@lanl.gov>]

I see in errno.h that the "errno" displayed periodically each time that
the CA server attempts to send a beacon is ENETDOWN (the network is
down).

#define	ENETDOWN	62		/* Network is down */

Otherwise, it is difficult to draw any definite conclusions. A stack
trace from the suspended "CA UDP" thread might help to further diagnose
the problem. A stack trace can be obtained by typing "tt <task id>", or
in this case "tt 0x7d9d260".

My best guess based on this information is that the CA UDP thread is
suspended because of vulnerabilities in this early release of EPICS
R3.13 to badly damaged UDP frames. If many bits are damaged in the
incoming UDP frame then the Ethernet CRC checksum and the UDP CRC
checksum may not detect the damage and an invalid message might be
passed through into the CA server software without its being detected. I
made a patch to an R3.13 release (after R3.13.1) to improve the
robustness of the CA server in an environment where incoming frames are
invalid. This was in response to problems first seen on a network where
the 10/100 auto negotiation was not working correctly if the network
switch was powered up after the IOC.

I am suspicious that this "network is down" message is an indicator for
the root cause of your problem. It seems that either the vxWorks network
interface driver or the vxWorks IP kernel has failed. This could be
because some errant code has damaged the data structures of these
components or because these components have experienced internal
failures. I would make certain that you are running with all available
WRS patches for your vxWorks distribution and also for your Ethernet
interface driver. Specifically, I recall having problems with the
initial version of WRS's 3Com 10/100 Ethernet interface card driver
during regression testing. After I installed all of the vxWorks patches
these problems vanished.

In summary, my recommendation is to upgrade to the latest patch levels
for EPICS R3.13, vxWorks, and the vxWorks driver for your network
interface card.

Jeff

> -----Original Message-----
> From: gfliu@ustc.edu.cn [mailto:gfliu@ustc.edu.cn]
> Sent: Monday, September 30, 2002 5:08 AM
> To: tech-talk@aps.anl.gov
> Subject: ask for help about ca
>
> Hi, all,
>
> epics has run successfully for more than two years on our x86 based
ioc
> at
> National Synchrotron Radiation Lab(epics:3.13.1, VxWorks:5.3.1).
> Recently
> we added a IOC (total 7 IOCs), and get the following error after the
IOC
> run about 2 days:
>
> Page Fault
> Program Counter: 0x001314c0
> Status Register: 0x00010206
> Error Code:0x00000000
> Task: 0x7d9d260 "CA UDP"
> filename="../taskwd.c" line number=175
> Task: 0x7d9d260 CA UDP suspended
>
> When I rlogin the IOC from host and try to reboot the IOC using "ctlr
X",
> the folling error is displayed on the IOC:
>
> 0x7f7b174 (CA online): ../online_notify.c "ca beacon error was "errno
=
> 0x3e"
>
> Could anybody give me some idea about it?
>
> Thanks in advance.
>
> Gongfa Liu
>