Subject: RE: does 3.0 support the smp on sparc64?
To: None <port-sparc64@NetBSD.org>
From: Andy Holt <andyh@andyh-rayleigh.freeserve.co.uk>
List: port-sparc64
Date: 12/31/2005 15:39:40
First attempt to send to list sent to wrong address, sorry ...


>  Now that 3.0 is out, I tried to find anything aboot SMP support in the
docs,
>  but I wasn’t able to.  Does that mean it’s not there yet?

I was going to raise this subject ...

I have tried compiling a MP kernel for an Ultra 60
It compiles without problems but:
During boot-up when the kernel tries to start the second processor
everything dies - just
locks-up with nothing responding.
If I compile the kernel for MP but don't allow it to find a second processor
it seems to behave just like a single-processor kernel.

A message on this list about a month ago implied that MP isn't yet working
... I seem to have confirmed this :-(
I understand that bug-fixing on the new console code has had priority (and
that seems reasonable); I hope that action can now resume on MP bug-fixing.

The GENERIC.MP config file really ought to have a "not working yet"
comment - and the documentation should also indicate this.

Incidently, enabling "ath" or "ray" in the configuration leads to a failed
compilation.
(again, neither has the XXX warning comment that such a problem should
deserve)

Note to the developers: I am willing to do a certain amount of testing ... I
have both a U2 and a U60 with two processors each - also have a U10 and a
U30 for that matter - and a variety of interface cards. However, I only have
a fairly slow net connection at the moment (ISDN - 64K, sessions limited to
2 hours) so cannot do large downloads ... and that is on a Windows 2000
machine, not directly to the Ultras.

Andy