Experimental Physics and
Industrial Control System

[email protected] (Bill Brown) · Wed, 18 Jun 1997 08:06:35 -0700

Oddly enough, we've had a similar situation with "a system that worked
fine 5 years ago.  We got it out of the closet and now it doesn't work."

I haven't looked into it closely yet, but I've heard that in the course
of recovering the disk the system boots from, the version of vxWorks used
got changed from vx5.1.1 to vx5.2.

WRS makes some unfortunate assumptions (at least from our perspective) about
what the default memory mapping should be for boards with various amounts
of memory installed.

If you're '147 has more than 4 megs of memory installed, it may well be the
case that none of it is being mapped to the A24 address space.  (Note:  the
board must be configured as processor #0 for _any_ memory to be mapped to the
vme bus.)  It may even be that it's broken for 4 meg boards - I just don't
remember for sure.  In any case, the '1014 must see the cpu memory in A24
data space.

I vaguely recall 'way back when the '147 boards were new - this was in our
"pre-EPICS" days, vx5.0 or 5.1 - I had to tweak the BSP because of a similar
problem. I haven't used '147 cpus in probably 5 years, so I don't remember
all the details.

I think the memory<->vme mapping gets set up in sysProcNumSet() which is in
sysLib.c.  In order to untangle what's doing which you'll need to have the
'147 hardware manual and untangle the way the VME-II chip works.  It's not
trivial, but it is do-able.  (SHUDDER!)

>    If the 1014 was installed properly, then its not a simple problem. When we
>    were porting EPICS to the MVME167, John Winans had to patch the
>    board support package (BSP) to make A24 address space available to DMA
>    devices.

He did a similar patch for the '162 also.
>	 It seems likely that the MVME147 BSP would require a similar patch.

I think both cpus ('147 & '167 use the VME chip, altho I can't be sure that
there isn't more than one version of it, and I'm not sure that the registers
are used in the same manner.  Perhaps a comparison of the appropriate parts
of the sysLib.c files may be helpful.
>    Since John is no longer around, you may have to plead with the greater
>    EPICS community for someone who has done this for the MVME147.

I hope this is at least somewhat helpful, and not too far from reality.

Disclaimer:  Any opinions are my own and have	    |  -bill
    nothing to do with the official policy or the   |   [email protected]
    management of L.B.N.L, who probably couldn't    |   Berkeley, CA
    care less about employees who play with trains. |   aka [email protected]

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System