Re: [patch] x86: make UV support configurable

[Ingo Molnar <mingo@elte.hu>]

* Jack Steiner <steiner@sgi.com> wrote:

> On Tue, Jan 20, 2009 at 04:50:00PM +0100, Ingo Molnar wrote:
> > 
> > * Ingo Molnar <mingo@elte.hu> wrote:
> > 
> > > * Nick Piggin <npiggin@suse.de> wrote:
> > > 
> > > > Fixed the EFI omission since the last patch. I have not changed the 
> > > > Kconfig to include (what I think are) false dependencies (which really 
> > > > frustrate me when things get badly tangled up). If people want to test 
> > > > as much stuff as possible even that is not supported by their hardware, 
> > > > we have allyesconfig/allmodconfig. I don't think UV is particularly 
> > > > special in that respect (we allow Intel, AMD, etc CPUs to be configured 
> > > > out)
> > > > 
> > > > If anything, X86_UV should *select* MAXSMP, rather than the other way 
> > > > around. That would be more logical, but even then I don't like adding 
> > > > still a false dep because one might have a small UV system, or want to 
> > > > test with different NR_CPUS.
> > > > 
> > > > Anyway, I think this patch is good as-is. If someone wants to streamline 
> > > > config options with subsequent patches, that's great. If Ingo can be 
> > > > convinced of a better arrangement, fine.
> > > 
> > > Applied to tip/x86/uv, thanks Nick!
> > 
> > hm, build failure with the attached config:
> > 
> > 
> > drivers/built-in.o: In function `gru_cpu_fault_map_id':
> > : undefined reference to `per_cpu____uv_hub_info'
> > drivers/built-in.o: In function `gru_fault':
> > : undefined reference to `per_cpu____uv_hub_info'
> > drivers/built-in.o: In function `gru_try_dropin':
> > grufault.c:(.text+0xbfc6e): undefined reference to 
> > `per_cpu____uv_hub_info'
> > grufault.c:(.text+0xc0042): undefined reference to 
> > `per_cpu____uv_hub_info'
> > [...]
> > 
> > (if possible please send a delta fix on top of your previous patch.)
> 
> The GRU driver makes sense only on UV & IA64 configurations.

that's not expressed in the Kconfig rules.

	Ingo