Re: 2.1.80 no-go

Erik Andersen (
Wed, 21 Jan 1998 12:46:28 -0700 (MST)

Umm. How about we make CONFIG_SMP a config option, and make a new
CONFIG_IO-APIC_SMP config option as well. After all, we now have the nice
new smart-config in the kernel, which means the reason CONFIG_SMP was
moved to a Makefile is no longer an issue. I believe all the
pre-conditions you previosly listed for changing this back are now in
place. I believe that Michael Elizabeth Chastain <> has
already written such a patch that works within the new smart-config
stuff. Linus, any chance of seeing this change soon?


Erik B. Andersen   Web: 
--This message was written using 73% post-consumer electrons--

On Wed, 21 Jan 1998, Linus Torvalds wrote:

> > > On Tue, 20 Jan 1998, Bill Hawes wrote: > > > > Kernel 2.1.80 doesn't boot for me -- I get keyboard timeouts, then > > timeouts on hda, hdb, etc. The results with and without the pirq= boot > > command seem to be about the same. On my system the pirq command gives > > > > pirq=0x0a,0x0b > > > > The system is a Pentium 133, Asus p55t2p4 w/ 32M, EIDE HD. > > Right. The current io-apic code probably doesn't work at all on non-SMP > systems, so you have to compile for UP to make it work. > > I'll probably make it default to using the old setup, with a run-time > utility to set up the IO-APIC interrupt handling. That way we won't be > making any mistakes at bootup, so the machine will at least work, and we > can write a small user-level program that parses the PCI irq tables to > find out what the pirq setting should be.. > > That should also make the SMP kernel boot on an UP machine again. > > Linus > >