On Tue, 1 Aug 2000, Simon Wing-Tang wrote:
> I'm attempting to get Linux to come up SMP on an old NCR 3450
> server. This server has 8 Pentium 90 CPU's and runs the old MCA bus
> architecture. The details I know so far are as follows....
That sounds like an interesting configuration.
> * System - NCR 3450
> * CPU - 8X90MHZ
> * MEM - 1024MB
> * NCR certified MP1.1/1.4 compliant
It should run Linux MP then but adjustments might be needed.
> * NCR certified NT4.0 ( with vendor supplied "HAL" )
> * System boots Linux using the Slackware ibmmca.s boot image ( which
> is SMP aware )
> * System reports " Jul 27 14:48:00 aardvark kernel: SMP motherboard
> not detected. Using dummy APIC emulation."
Hmm, it looks like an MP floating pointer could not be found. What
version of Linux are you running? I would recommend trying the newest
development kernel first -- try 2.4.0-test5 which is the latest
prerelease. There were serious updates for systems like yours recently.
> I'm guessing that the kernel does a probe of the BIOS to get the SMP
> information it requires. On some motherboards the BIOS needs to be set to
The kernel scans certain predefined memory areas for the so-called MP
floating pointer. That's a small array that indicates the system is
MP-compliant, provides the revision of the MPS the system adheres to and
optionally provides a pointer to an MP table that describes the system in
details.
> MP1.1/1.4 mode or Unixware or something like that. There is no option to
> set anything like this in the 3450 as it was built to run NCR MP-RAS UNIX.
Hmm, look carefully -- the option may be named strangely sometimes.
> Under the vendor OS it sees all 8 CPU's with no problems, but not under
> Linux.
How does the OS describe the system? It should give some notes on the
configuration. If the system is to be MP-compliant it has to provide an
MP floating pointer in one of the predefined areas.
> So, is there anyone else out there that has Linux running on one of
> these machines and managed to get SMP going ? I'm also guessing that it
> could be possible to force the SMP probe to succeed and manually set up the
> kernel with the required parameters, this may be more effort than it's worth
Forget it -- the configuration of interrupts may change across boot-ups.
> though. Any ideas here? By the way, I have checked the usual sites for SMP
> and MCA Linux but have found nothing to help so far.
The fixes for handling of MCA interrupts in APICs went in to one of
2.4.0-test* kernels. I couldn't test them on any real hardware, so they
are only hopefully correct by specs.
--
+ Maciej W. Rozycki, Technical University of Gdansk, Poland +
+--------------------------------------------------------------+
+ e-mail: [EMAIL PROTECTED], PGP key available +
-
Linux SMP list: FIRST see FAQ at http://www.irisa.fr/prive/dmentre/smp-howto/
To Unsubscribe: send "unsubscribe linux-smp" to [EMAIL PROTECTED]
