Reset NVRAM on DECstation 5000/240

Maciej W. Rozycki macro at linux-mips.org
Tue Aug 1 13:37:37 CDT 2017


On Mon, 31 Jul 2017, Rob Jarratt wrote:

> >  It is.  Check the "DECstation/DECsystem 5000 Model 200 Series Maintenance
> > Guide" (EK-PM38C-MG-002), available online.
> > 
> 
> I have that manual. Having checked again I realise I should have searched
> for "NVR" and not "NVRAM", then I would have found it!!

 Or "jumper" would do too, ;) avoiding DEC documentation peculiarities 
(they indeed call the Dallas chip "NVR", reserving "NVRAM" for the MS02-NV 
Prestoserve SRAM memory board).

> >  Technically it's wired to one of the 16 IOASIC external interrupt inputs,
> which is
> > otherwise unused.  Functionally it tells software whether the mainboard is
> used
> > in a 5000 or a 5900 system configuration.  I can dig out the exact details
> if
> > you're interested.
> 
> 
> That would be interesting, if there is documentation that isn't generally
> available it would be good to see it.

 Not everything has been properly documented, I've gathered information 
from various sources, including reverse engineering.  The bigmax IOASIC 
IRQ bit is I recall only mentioned in an Ultrix error log processing 
utility, though the jumper itself is I believe mentioned in one of the 
5900 manuals, available online.

 Over the years I've gathered available information in Linux kernel 
headers; for the KN03 mainboard specifically:

<https://git.linux-mips.org/cgit/ralf/linux.git/tree/arch/mips/include/asm/dec/kn03.h>

Interpreted along with:

<https://git.linux-mips.org/cgit/ralf/linux.git/tree/arch/mips/include/asm/dec/ioasic_addrs.h>

and knowing that in the console you need to use the MIPS KSEG1 uncached 
address space you get:

KSEG1 | (KN03_SLOT_BASE + IOASIC_IOCTL + IO_REG_SIR)

which expands to:

0xa0000000 | (0x1f800000 + 1 * 0x00040000 + 0x110) => 0xbf840110

so to get at the interesting bits:

>>e -wx 0xbf840110
bf840110: 0x00008802
>>

i.e. the bigmax bit is set as for the 5000/240 (jumper present), the NVR 
clear bit is clear (jumper absent), and the NRMOD bit is clear (jumper 
absent).  There is also a TURBOchannel slot #0 option interrupt pending 
here for some reason and the halt button is released.

> >  BTW I saw your 5000/240 web page and I'm glad that you have rescued all
> > that fine hardware.  Also I think you don't have to worry about the H7878
> PSU --
> > apparently none of the caps used in that model suffers from the dreaded
> > quaternary ammonium salt issue.
> 
> 
> Thanks. I will create a page describing the machine in the coming days, for
> now it will just be the blog posting. I do think there is a problem with the
> PSU though because it will be just sitting there in the console and power
> off, you can't turn it back on until you pull out the power cord. I am going
> to check the caps more closely.

 Always worth doing of course, although from its arrangement the H7878 
would be a pain do disassemble to reach some of its caps.  Also check for 
the PSU overheating condition, reported at bit #4 in the IOASIC interrupt 
status as examined above (with good status indicated with the bit clear).  
It may cause the PSU to eventually shut itself down.

  Maciej


More information about the cctech mailing list