On Fri, Feb 07, 2014 at 03:09:09PM -0800, David Rientjes wrote:
> On Fri, 7 Feb 2014, Andrew Morton wrote:
>
> > > > > > diff --git a/include/linux/mm.h b/include/linux/mm.h
> > > > > > index 1cedd00..5f8348f 100644
> > > > > > --- a/include/linux/mm.h
> > > > > > +++ b/include/linux/mm.h
> > > > > > @@ -1589,10 +1589,8 @@ static inline int __early_pfn_to_nid(unsigned long pfn)
> > > > > > #else
> > > > > > /* please see mm/page_alloc.c */
> > > > > > extern int __meminit early_pfn_to_nid(unsigned long pfn);
> > > > > > -#ifdef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
> > > > > > /* there is a per-arch backend function. */
> > > > > > extern int __meminit __early_pfn_to_nid(unsigned long pfn);
> > > > > > -#endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
> > > > > > #endif
> > > > > >
> > > > > > extern void set_dma_reserve(unsigned long new_dma_reserve);
> > > > >
> > > > > Wouldn't it be better to just declare the __early_pfn_to_nid() in
> > > > > mm/page_alloc.c to be static?
> > > >
> > > > Won't that break the ability to override that function in
> > > > architecture-specific code (as arch/ia64/mm/numa.c does)?
> > > >
> > >
> > > Why? CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID should define where this function
> > > is defined so ia64 should have it set and the definition which I'm
> > > suggesting be static is only compiled when this is undefined in
> > > mm/page_alloc.c. I'm not sure why we'd want to be messing with the
> > > declaration?
> >
> > __early_pfn_to_nid() must be global if it is implemented in arch/.
> >
>
> Why?? If CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID then, yes, we need it to be
> global. Otherwise it's perfectly fine just being static in file scope.
> This causes the compilation unit to break when you compile it, not wait
> until vmlinux and find undefined references.
>
> I see no reason it can't be done like this in mm/page_alloc.c:
>
> #ifdef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
> extern int __meminit __early_pfn_to_nid(unsigned long pfn);
No, a .c file should not have an extern declaration in it. This should
live in an appropriate header file, to be included in both page_alloc.c
and any arch file that defines an overriding function.
> Both of these options look much better than
>
> include/linux/mm.h:
>
> #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
> !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
> static inline int __early_pfn_to_nid(unsigned long pfn)
> {
> return 0;
> }
> #else
> /* please see mm/page_alloc.c */
> extern int __meminit early_pfn_to_nid(unsigned long pfn);
> #ifdef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
> /* there is a per-arch backend function. */
> extern int __meminit __early_pfn_to_nid(unsigned long pfn);
> #endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
> #endif
>
> where all this confusion is originating from.
The proposal is to first simplify those ifdefs by eliminating the inner
one in the #else; I agree with Andrew that we ought to go ahead and take
that step given the patch at hand, and then figure out if there's an
additional simplification possible.
- Josh Triplett
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@xxxxxxxxx. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>