Experimental Physics and
Industrial Control System

Andrew Johnson <[email protected]> · Fri, 21 Mar 2014 15:37:11 -0500

Hi Matt,

On 03/21/2014 02:21 PM, Matt Rippa wrote:
> I have an MVME6100 with Epics 3.14.12.2 running vxWorks 5.5.
> Sometimes when testing changes to a new vxworks kernel or just
> an update to a new epics module, my shell stops hard, and Ctrl-x
> doesn't work. Power cycling is my best friend in that case, but
> is it my only option?

Does hitting the VME Chassis reset button rescue you from those hang-ups?

> Can a modified reboot handler help with any of these occurrences?

I think by the time the CPU has frozen it's going to be too late, unless
a VMEbus reset does work and you can arrange for some way to trigger it
remotely.

Here at the APS we connect to most of our VME console ports through a
locally-designed adapter board which monitors the serial line for a
particular key sequence (usually Ctrl+Y three times in a row). When it
sees that, it pulses the VMEbus reset line, which is usually enough to
restart most frozen systems. If we're unlucky we might have to send
someone to walk out to the crate to power-cycle it, but that's fairly rare.

The other thing I would recommend is that you examine all of your
drivers for cards that use interrupts, and make sure that they all
register a reboot hook routine (or an epicsAtExit one) that disables the
interrupt on the card. In my experience the main reason for systems
freezing over a reboot is that they leave an interrupt line asserted,
and that prevents the OS from coming back up properly.

HTH,

- Andrew
-- 
Advertising may be described as the science of arresting the human
intelligence long enough to get money from it. -- Stephen Leacock

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System