Experimental Physics and
Industrial Control System

Hello all
I have been following this thread with great interest. We have been
struggling with the problem for quit sometime. It began when we upgraded
from vxWorks5.3 to 5.4. It was particullarly bad with iocs that use
the Intel 82559 LAN controller. Sometimes the bootrom code crashed
before we even got as far as running vxWorks let alone EPICS. A later
upgrade to 5.5.1 improved thing.

We now run all our iocs behind firewalls. However we still had one IOC
(an mvme162) that continued to experience CA connection problems
("S_errno_ENOBUFS"). It maintains alot more networks connections
than any other ioc. We have been forced to revert to vxWorks 5.3
This ioc now chugs along merrily even when all hell is breaking lose
around it.

Why Wind River changed from the dynamic allocation of network buffers
in 5.3 I will never understand.

Dear Andrew,

This problem is becoming critical.  Every vxWorks/PowerPC IOC at the
NSLS at Brookhaven is crashing within minutes if it is connected to the
NSLS network.  The only way the beamlines can run is to disconnect their
switches from the NSLS network.  Tim Mooney told me yesterday that
several CATs are now seeing this problem here as well.  It is not known
if the network activity that is crashing the IOCs is a malicious or not.
Finding the source of the packets at the NSLS is being worked on, but
the real solution is to see if there is a fix for the network software
on vxWorks to make it immune to this problem.  Can you see if Wind River
has a patch that fixes this?

Thanks,
Mark

--
+---------------------------------------------------------------------+
|Graham Waters, Control System Eng    phone: (604)-222-1047 ext 6531  |
|TRIUMF --- University of B.C,        Fax: (604)-222-7307             |
|Vancouver, BC, Canada                e-mail: [email protected] |
|                                                                     |
|http://isacwserv.triumf.ca/homepage/graham01.html                    |
+---------------------------------------------------------------------+

References:

RE: Buffer problems Mark Rivers

Navigate by Date:

Prev: Re: mvme5500 (was National Instruments VME-MXI-1 modules vs. modernVMECPUmodules) Eric Norum

Next: RE: Buffer problems Ernest L. Williams Jr.

Index: 1994  1995  1996  1997  1998  1999  2000  2001  2002  2003  <2004>  2005  2006  2007  2008  2009  2010  2011  2012  2013  2014  2015  2016  2017  2018  2019  2020  2021  2022  2023  2024 

Navigate by Thread:

Prev: RE: Buffer problems Mark Rivers

Next: RE: Buffer problems Ernest L. Williams Jr.

Index: 1994  1995  1996  1997  1998  1999  2000  2001  2002  2003  <2004>  2005  2006  2007  2008  2009  2010  2011  2012  2013  2014  2015  2016  2017  2018  2019  2020  2021  2022  2023  2024