Experimental Physics and
Industrial Control System

"Leonard J. Reder" <[email protected]> · Fri, 15 Oct 1999 09:46:06 -0700

Jeff,

I am just now getting my first Tornado2 kernel to build and am planning
to give EPICS a try on it next week.  It has been a bit of a struggle
for me due to networking problems.  It seems that the standard Tornado2
kernel must auto-negotiate with the network and from certain places
on lab this does not seem to work.  Wind River has a patch to allow the
auto negotiation to be turned off . I will be trying this patch and then
if all goes well should be able to test EPICS on Tornado2.

We already have Tornado2 running in one of our labs with ACE/TAO (e.g.
real time CORBA support) and our own internal software package.  It
is highly desirable for us to have EPICS on Tornado2 since we must
be running ACE/TAO on Tornado2 anyway.  It would be preferred to just
use the newer Kernel rather then switching back and forth which I have
been doing for about a month now.

Just to refresh your memory I am working on the Keck
Interferometer project which will be installed at the Keck observatory
starting next year.

So EPICS supported on Tornado 2 would be a really good thing and I will
try and
help iron out the bugs with this as best I can.  I encourage you to
keep working on this.

Thanks,

Len

Jeff Hill wrote:
> 
> Markus,
> 
> Here are the known problems integrating the EPICS with the WRS SENS
> IP kernel (which appears to be mandatory in Tornado II):
> 
> 1) There is a SENS supplied m_type macro which collides
> with the CA source codes. This is fixed in 3.14.
> 
> 2) There are problems querying the available network interfaces using
> the socket library. This causes the CA client library to break
> unless additional (non-standard) configuration steps are taken.
> This also causes the CA server to neglect to send beacons
> unless additional (non-standard) configuration steps are taken.
> I have a TSR outstanding with WRS for more than a month on this.
> 
> 3) There is an initialization order problem which causes trouble
> when a messages about (2) is attempted. This is fixed in 3.14.
> 
> 4) During high throughput load testing of Channel Access I have been
> experiencing SENS IP kernel lock ups. This appears to be related to
> buffer space starvation, but I am unable to reliably fix the problem
> when I increase the buffer space available to the IP kernel. So far,
> I don't see anything that we are doing to cause this problem.
> 
> You may be able to work around (2) by setting EPICS_CA_AUTO_ADDR_LIST=NO
> and then configuring EPICS_CA_ADDR_LIST to the broadcast address of
> the network interface in your IOC. This can be determined by typing
> "ifShow" at a vxWorks shell prompt.
> 
> How many sites are waiting for EPICS to be made available under
> Tornado II?
> 
> Jeff
> 
> >
> > there has been an e-mail about the new "SENS" IP kernel from Wind
> > River Systems on tech-talk dated December 7, 1998 by J. Hill. In there
> > he recommended to not specify the "SENS" option when building Tornado.
> > This has the drawback that WindView is not available.
> >
> >  The problem is that there is a name clash for the PowerPC
> > architecture. WRS defines in one of their header files (net/mbuf.h) a
> > macro m_type which clashes with a variable of the same name in a
> > structure used by Channel Access. I changed the name of the variable
> > m_type in the code of CA (10 files had to be changed) and now CA seems
> > to work with the SENS IP kernel!
> >
> >  The files that were changed are given below. The path is relative to
> > EPICS_BASE/src/
> >
> > ca/access.c
> > ca/caProto.h
> > ca/service.c
> > cas/generic/casClient.cc
> > cas/generic/casClientMon.cc
> > cas/generic/casCtxIL.h
> > cas/generic/casDGClient.cc
> > cas/generic/casStrmClient.cc
> > cas/generic/outBuf.cc
> > rsrv/camessage.c
> >
> >  I tested the new EPICS with the new kernel on a couple of IOC
> > applications and discovered, so far, no new problems using EPICS
> > 3.13.0.beta12, Linux as host and mv2306 as target. A test on solaris
> > and mv167 will be done next week, but I don't expect any
> > problems. Could the above mentioned changes included later into the
> > EPICS distribution? Jeff?
> >
> >  Cheers, Markus
> > --
> > +---------------------------+--------------------------------+
> > | Markus Janousch           | e-mail: [email protected] |
> > | Computing & Controls, SLS |                                |
> > | Paul Scherrer Institute   |    Tel: +41-56-310 3275        |
> > | CH-5232 Villigen-PSI      |    Fax: +41-56-310 4413        |
> > +---------------------------+--------------------------------+
> >
> >
> >

-- 
____________________________________________________
Leonard J. Reder
Jet Propulsion Laboratory
Interferometry Systems and Technology Section 383
Email: [email protected]
Phone (Voice): 818-354-3639
____________________________________________________

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System