Experimental Physics and
Industrial Control System

Andrew Johnson <[email protected]> · Wed, 20 Jan 2010 13:46:12 -0600

On Wednesday 20 January 2010 13:27:47 [email protected] wrote:
>
> I just tried it on a VME5100 (instead of VME2700) in a test crate and it
> was picked up fine using the same Ipac.  So now it's at least isolated to
> this IOC hardware and additional software.  What's still weird is that
> doing m(0xfbff0000,2) on the problem IOC shows all memory set to ffff.
>
> vxwrks@iocfel7 -> m(0xfbff0000,2)
> fbff0000:  ffff-
> fbff0002:  ffff-
> fbff0004:  ffff-
> fbff0006:  ffff-
> fbff0008:  ffff-
> fbff000a:  ffff-

With some BSPs as delivered by Wind River a bus error from a VME read cycle 
can give all ones data just like that, without raising any kind of exception.  
You should find that you get the same result in any of the VME Master window 
address ranges.  With these BSPs using vxMemProbe() is the only way to tell 
whether something is responding at a particular address.

I have had to make changes to most of the PowerPC BSPs used at APS so that a 
bus error will cause an exception like we're used to from 68K family CPUs, 
although with the write FIFOs found in modern CPUs and bus bridges it is 
usually impossible to reliably work out which instruction caused a bus error 
on write (reads work OK though).

- Andrew
-- 
The best FOSS code is written to be read by other humans -- Harald Welte

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System