On Thu, May 14, 1998 at 06:00:41PM -0400, Anderson MacKay wrote:
> I'm think that if you have a uniprocessor kernel running, it's rather
> difficult to tell whether the hardware supports SMP. Now, it's possible
> to make the default kernel an SMP kernel ... as far as I know, this is how
> Linux compiles "out of the box" in 2.1.x. If you don't go in and comment
> out "SMP=1" in the Makefile, it builds SMP. On uniprocessor machines it
> doesn't do anything different than normal, it runs just the same. But
> when you put in on an SMP box, off it goes dual or quad proc. If it's
> possible to compile an SMP kernel and boot it with the kernel option to
> tell it that there's only one CPU, that may be the best of both worlds.
> It keeps the kernel from possibly being confused on SMP-capable boards
> with only one processor installed, but I think ... heck, I dunno. But I
> would guess that then the multi-CPU info is at least available in /proc.
> Generally, though -- if I'm not mistaken -- an SMP-capable kernel should
> be able to autodetect how many CPUs there are and adjust accordingly. So
> making the default kernel an SMP-aware one probably isn't going to cause
> problems, should you decide to do that.
You are right about the above.
What i will do is to:
1) package a kernel-image for SMP so that SMP users won't have to
compile their own kernel and just have to install it with dselect.
2) Get info on own to detect if possible a SMP MB on a UP kernel.
3) See what kind of feedback, i get from the discussion so far.
Regards,
--
Eric Leblanc -- jughead@generation.net -- Be happy use GNU/Linux
As for the M$ pundits claim that the problems result from an
incorrect setup/configuration: they are perfectly correct, if you
hadn't installed an M$ OS, then the crashes wouldn't be happening.
-- Jeff Dutky on c.o.l.a
Attachment:
pgpHx4mhDK1UG.pgp
Description: PGP signature