Subject: SUMMARY: Re: Isolating memory error
To: None <port-sparc@NetBSD.ORG>
From: Greg Earle <earle@isolar.DynDNS.ORG>
List: port-sparc
Date: 12/10/2003 12:42:07
'Way back three months ago, der Mouse said:
>>> On a 20, SIMMs appear 64MB (the max SIMM size) apart.  Their base
>>> addresses are multiples of 04000000.  The address you cite,
>>> 0x01e76b89c, is 7*0x04000000 + 0x276b89c.  Thus, counting SIMMs from
>>> 0, it is in SIMM number 7 (which is obviously a 64M SIMM, since it
>>> contains an address above the 32M mark).
>>
>> Thanks for the tip on turning on diag-switch?.  It says:
> 
>> Probing Memory Bank #0 64 Megabytes of DRAM
>> Probing Memory Bank #1 64 Megabytes of DRAM
>> Probing Memory Bank #2 Nothing there
>> Probing Memory Bank #3 Nothing there
>> Probing Memory Bank #4 Nothing there
>> Probing Memory Bank #5 64 Megabytes of DRAM
>> Probing Memory Bank #6 Nothing there
>> Probing Memory Bank #7 32 Megabytes of DRAM
> 
>> Do you guys still think it's in bank #7 now, given that #7 contains a
>> 32 Mbyte SIMM and not a 64 Mbyte SIMM, as you previously had surmised?
> 
> No.  Now I don't feel at all confident of the analysis above.
> 
> It could be that the 64M stride I gave is wrong.  But 128M would put it
> in bank 3, which is even less plausible.  And if SIMMs are addressed
> 256M apart, the address given is in SIMM 1 but at an offset well over
> 64M from the SIMM's base.
> 
> In my recent experience, which is with a 10 rather than a 20, I also
> see "testing memory" lines that give base addresses and count down the
> sizes.  If you get anything of the sort (my 20 is not in a circumstance
> where it can be casually power-cycled), that can give you the memory
> area bases and sizes.

Just wanted to post a long-overdue summary/update:

I upgraded from 1.4.1 to 1.6.1 over the past couple of days (more on
that later - the usual NetBSD upgrade/install Hell is still present),
and in rebuilding a lot of my old pkgsrc stuff, the memory error
started driving me bananas.

I took out the 32 Mbyte SIMM in slot 7 and that cured the problem.
Not a single memory error since then.

While I don't like running with 192 Mbytes instead of 224, getting
rid of the memory errors was worth it!

Thanks to der Mouse and Julian Coleman for their help and assistance.

	- Greg