number of CPUs and IPI panic

[Josh Tolbert]

On Thu, Nov 11, 2004 at 05:02:43PM -0500, Ketrien I. Saihr-Kesenchedra wrote:
> 
> To answer some of your questions Stephan;
> The ALR Revolution 6x6 is a 6-way PPro; it uses two processor cards and 
> a SIMM or DIMM memory card. The processor cards contain partial BIOS, 
> VRMs, and CPUs (obviously.) They come in 1 through 3-way flavors. These 
> will -always- label the last processor as BSP; this is a hack so that 
> the installation socket is irrelevant. (ie; I can have one CPU in the 
> 3rd socket of a 3-way board, and the system will boot.) The system in 
> question here is undoubtedly a Unisys Aquanta HS/6 or HR/6. A genuine 
> ALR has no mention of Unisys in MPTable. The board has two distinct PCI 
> buses in addition to the EISA bus, which we do not appear to be 
> detecting correctly. In addition, we appear to erroneously detect a pcic 
> device; this may be the keyboard/mouse card. (Please forgive, my memory 
> is somewhat fuzzy.)
> The amount of logic on the motherboard is nothing short of staggering; 
> the core is indeed i440GX Orion. HOWEVER, take a look at this picture:
> http://www.vanvleet.net/images/rev6x6bd.gif
> To the left of the single shared PCI/EISA slot are the PCI controllers. 
> Between them is a crystal, and above is a NatSemi SuperI/O (I believe.) 
> BOTH controllers are FULL southbridge chips. As you look between the CPU 
> slots and Memory slot, you should find two FULL i440GX's. That's not a 
> mistake; part of the ALR's magic is that the maximum number of PPros on 
> a chipset is 4; to get around this, they used two. I tried to find a 
> better picture to be more certain of the identification, but no such 
> luck I'm afraid, and the ALR's I have access to are running and can't be 
> disassembled.
> 
> We have a very big problem here though; we're detecting 7 processors. 
> The ALR has a LOT of on-board logic. Handling it is very tricky; these 
> things required a special version of Windows NT or SCO, as I recall. We 
> should only be detecting 6 CPUs; we may be erroneously accepting part of 
> the bridging logic as a processor. (ISTR that the CPU board bridge 
> appears very similar to a processor.) I think an mptable dump would be 
> very helpful here; I'm wondering if we're not letting the board fake us 
> out with bridge logic presenting as a processor.
> 
> -Ketrien I. Saihr-Kesenchedra / ketrien at error404.nls.net
> _______________________________________________
> freebsd-current at freebsd.org mailing list
> http://lists.freebsd.org/mailman/listinfo/freebsd-current
> To unsubscribe, send any mail to "freebsd-current-unsubscribe at freebsd.org"

Win2K Advanced Server works just fine on these boxes, no special patches
needed. FreeBSD-4.x ran just fine on my Revolution 6x6 (also 6x PII OverDrive;
you can see it at http://www.puresimplicity.net/~hemi/Pics/Comps/ALR/). If you
want high-detail pics of a 6x6 motherboard, let me know. I have a spare dead
one that I looted some parts from to revive my working one.

I haven't tried FreeBSD-5.x on my 6x6, mainly cause the thing is disassembled
until I build a case for it. Mine never had problems crashing under heavy load
with 4.x, though.

Thanks,
Josh
-- 
Josh Tolbert
hemi at puresimplicity.net  ||  http://www.puresimplicity.net/~hemi/

If your sysadmin's not being fascist, you're paying him too much.
   --Sam Greenfield