SMP i586?

[Lamar Owen]

On Wednesday 26 November 2003 11:54 am, Dave Jones wrote:
> For the curious, this is the dual P90 that Caldera[*] sent Alan
> for SMP development. It's incredibly clunky, weighs a ton, and
> in comparison to any of todays hardware, is almost painful to use
> for anything serious.

Sounds alot like the HP NetServer LS 5/133 I have.... big, bulky, heavy, and a 
royal pain to use (mixed EISA PCI box).  But it_was_ a freebie, and it works 
great, has a built-in aic7xxx dual channel controller, and a six bay hotswap 
SCA chassis.

It's being built up as a CD server, with 98 CD-ROM drives hooked up to it, 
eventually.  The CD's are rack mounts made by SMS and include the requisite 
LUN SCSI Expander with seven CD's per unit, attached via two AHA-1740's.  
Will be interesting to see if it works correctly with the Linux kernel LUN 
support (I had to rebuild the kernel for multiple luns anyway, so building 
for SMP wasn't that big of a deal).

> It would be fun to get Fedora on that monster some time though.

It took IIRC a mere 2 hours to upgrade this beast to FC1, from RHL 8.0.  The 
upgrade zonked the lilo config, and blew out the SMP kernel at the same time.  
Interesting upgrade, it was.  The reboot acutally picked up the old kernel (I 
had a custom 2.4.20 kernel with SuperFreeS/WAN, which was a good thing, 
otherwise the box might not have booted at all), but tons of module errors 
later I was able to edit lilo.conf, recreate some lilo files, and rerun 
/sbin/lilo to get the thing booting again.  And I know it's not a supported 
config....

> I thought we were actually churning out i586-smp kernels, but
> I just checked and spotted the exclusion in the spec file.
> I'll nuke that, and push one out sometime for folks to play with.

Dave, one day when you have time I'd like for you to explain the kernel spec's 
use of --with and --without and how that's communicated to the spec file 
build logic.  I would love to use --with and --without to do the conditional 
subpackage building in PostgreSQL instead of the more unweildy --define 'tcl 
0' etc.  This might even be interesting to others on list.
-- 
Lamar Owen
Director of Information Technology
Pisgah Astronomical Research Institute
1 PARI Drive
Rosman, NC  28772
(828)862-5554
www.pari.edu