Experimental Physics and
Industrial Control System

"GOURNAY Jean-Francois" <[email protected]> · Mon, 22 Jun 2009 11:27:52 +0200

Hello,

After some exchanges with WRS, it seems that this problem has already occurred for several clients, especially those using a WRS SBC which has the same INTEL strata flash memories as those present on the 5500:

MVME5500> flashShow
Device-Name  Base-Address,Size  Device-Size,Count  Boot  Type
/dev/flash0  F2000000,02000000  01000000,00000002  No    Intel 28F128
/dev/flash1  FF800000,00800000  00400000,00000002  Yes   Intel 28F320

I fully agree with Nick and Andrew that this problem can be avoided at least on flash0, which could be protected by S3-1 switch. The alternative is to protect either MOTLoad or VxWorks boot. It is not clear for us today what will be the better choice.

J.F. Gournay
CEA Saclay
IRFU/SIS

-----Message d'origine-----
De : Rees, NP (Nick) [mailto:[email protected]] 
Envoyé : vendredi 19 juin 2009 17:33
À : Andrew Johnson; [email protected]
Cc : GOURNAY Jean-Francois
Objet : RE: MVME 5500

Hi Jean-Francois,

We have 5500's running VxWorks and don't suffer from this problem.

VxWorks should be in Flash 0, and MOTLoad in Flash 1. Use J8 to select
which bank of flash to boot from. Jumper pin 1-2 for VxWorks and 2-3 for
MOTLoad.

Removing jumper J9 should disable writing to Flash 0, which is what
Andrew wanted to find out. However, I think we have J9 installed, so I
am a bit surprised at your problem. You may, however, a very specific
loose pointer.

Cheers,

Nick Rees
Principal Software Engineer           Phone: +44 (0)1235-778430
Diamond Light Source                  Fax:   +44 (0)1235-446713 

> -----Original Message-----
> From: [email protected] 
> [mailto:[email protected]] On Behalf Of Andrew Johnson
> Sent: 19 June 2009 15:21
> To: [email protected]
> Cc: GOURNAY Jean-Francois
> Subject: Re: MVME 5500
> 
> Hi Jean-Francois,
> 
> On Friday 19 June 2009 06:25:29 GOURNAY Jean-Francois wrote:
> >
> > Whenever a C program has a misbehaved pointer, which causes a
> > memory-write outside the allocated data memory, we have a 
> crash of the
> > CPU which is absolutely understandable, but then, it is 
> impossible to
> > reboot the CPU. Both flash banks containing the initial 
> bootload and the
> > VxWorks boot are corrupted and the only solution is to send 
> back the CPU
> > to the retailer !!! We had experienced this with several 
> CPUs (we are
> > probably bad C programmers) and didn't get so far any 
> explanation from
> > our retailers or after sales services. We have VxWorks 6.5.
> 
> Does this board have write-protect jumpers for the Flash?  
> Most of the other 
> MVME boards do, but I can't get to the hardware documentation for the 
> MVME5500 right now on the Emerson website so I can't check 
> that for myself.  
> If it does they should at least be able to prevent the flash 
> from getting 
> corrupted when things go wrong.
> 
> HTH,
> 
> - Andrew
> -- 
> The best FOSS code is written to be read by other humans -- 
> Harold Welte
> 
-- 
Scanned by iCritical.

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System