I had a NetFinity 7000 on loan until about 10 days ago, when I returned
it. Here's a quick summary of what I found:
- mono 2.0 and 2.2 kernels ran fine and were stable
- SMP 2.0 and 2.2 kernels could not be started at all via LILO, even
using the newest versions and adjusting the load point down in memory.
- I could consistently get SMP 2.0 and 2.2 kernels running with a "plain"
kernel floppy - i.e., dd your bzImage directly to /dev/fd0 and boot from
that floppy.
- SMP 2.0 kernels saw all four processors, but only 1 GB of memory.
Kernels suitably modified to see 2 GB would die in the middle of
booting.
- SMP 2.2 kernels - the last I tried was 2.2.5 - saw all four processors
at boot time. However, subsequent load tests showed processes running on
only 3 of the 4 (the smp-modified top which shows usage per CPU showed up
to 100% on 3 of the 4, 0 % on the other; further, scaling tests clearly
showed only 3 processor's worth of work could be done). These kernels
were stable for days at a time.
- SMP 2.2 kernels suitably modified would see all 2 GB of memory.
- I found a patch via dejanews to enable multiple APICS (note, I heard
this would not be supported until 2.3) and indeed 2.2 kernels booted fine
and distributed interrupts to all processors (under 2.2.5 this also fixed
the dormant 4th processor). However, such kernels typically died
silently after 15 to 30 minutes.
The Netfinity 7000 is based on Intel's AD450NX. I've since played with
quad systems based on the SC450NX and have had no problems whatsoever.
Don Holmgren
Fermilab
On Tue, 25 May 1999 [EMAIL PROTECTED] wrote:
> "Thomas G�tz" wrote:
>
> > Hello,
> >
> > one of our customers had an Netfinity 7000 with 2GB Memory and 4 Xeon-400
> > Processor. The System is not running very well with 4 Procs. The Uniprocessor
> > System runs fine. At first we asumed that the graphics could cause the locks
> > but this is not true. We don't run any special application and the computer
> > compile the kernel without problems but if you do nothing the kernel locks
> > after a few minutes.
> >
> > I/O APIC #14 Version 17 at 0xFEC00000.
> > I/O APIC #13 Version 17 at 0xFEC01000.
> > Warning: Multiple IO-APICs not yet supported.
> > Warning: switching to non APIC mode.
> > Processors: 4
> > WARNING: MP table in the EBDA can be UNSAFE, contact [EMAIL PROTECTED]
> > if you experience SMP problems!
> > mapped APIC to ffffe000 (fee00000)
> > mapped IOAPIC to ffffd000 (fec00000)
> > Detected 400033054 Hz processor
> >
> > The maschine is currently not a production machine therefore we can test
> > configurations. I hope you could help us.
> >
> >
>
> I have also been working with the same machine trying to get it to work under
> linux.
> Can you write and tell what kernel version you are using, what distribution and
> what
> options you compiled your kernel with? I have been trying to make it work with a
> 2.2.x
> kernel - currently 2.2.6. I have never been able to get any of the 2.2.x kernels
> to detect
> APICs or more than one processor.
> The machine was previously running with 2.0.33 and experiencing lockups at regular
>
> intervals.
> It would be very nice to talk to someone with a similar machine and find out what
> you
> have done and compare notes.
>
> Brian Murphy
>
> -
> Linux SMP list: FIRST see FAQ at http://www.irisa.fr/prive/mentre/smp-faq/
> To Unsubscribe: send "unsubscribe linux-smp" to [EMAIL PROTECTED]
>
-
Linux SMP list: FIRST see FAQ at http://www.irisa.fr/prive/mentre/smp-faq/
To Unsubscribe: send "unsubscribe linux-smp" to [EMAIL PROTECTED]
