On Tue 04-07-17 14:11:41, Joonsoo Kim wrote:
> On Fri, Jun 30, 2017 at 05:44:16PM +0200, Michal Hocko wrote:
> > On Fri 30-06-17 17:42:24, Michal Hocko wrote:
> > [...]
> > > diff --git a/include/linux/mmzone.h b/include/linux/mmzone.h
> > > index 16532fa0bb64..894697c1e6f5 100644
> > > --- a/include/linux/mmzone.h
> > > +++ b/include/linux/mmzone.h
> > > @@ -1055,6 +1055,7 @@ static inline struct zoneref *first_zones_zonelist(struct zonelist *zonelist,
> > > !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP)
> > > static inline unsigned long early_pfn_to_nid(unsigned long pfn)
> > > {
> > > + BUILD_BUG_ON(!IS_ENABLED(CONFIG_NUMA));
> >
> > Err, this should read BUILD_BUG_ON(IS_ENABLED(CONFIG_NUMA)) of course
>
> Agreed.
>
> However, AFAIK, ARM can set CONFIG_NUMA but it doesn't have
> CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID and CONFIG_HAVE_MEMBLOCK_NODE_MAP.
$ git grep "config NUMA\|select NUMA" arch/arm
$
Did you mean arch64? If yes this one looks ok
$ git grep "HAVE_MEMBLOCK_NODE_MAP\|HAVE_ARCH_EARLY_PFN_TO_NID" arch/arm64/
arch/arm64/Kconfig: select HAVE_MEMBLOCK_NODE_MAP if NUMA
> If page_ext uses early_pfn_to_nid(), it will cause build error in ARM.
Which would be intentional if it doesn't provide a proper implementation
of the function.
> Therefore, I suggest following change.
> CONFIG_DEFERRED_STRUCT_PAGE_INIT depends on proper early_pfn_to_nid().
> So, following code will always work as long as
> CONFIG_DEFERRED_STRUCT_PAGE_INIT works.
I haven't checked all other callers of early_pfn_to_nid yet but I have
run my original patch (with !IS_ENABLED...) just to see whether anybody
actually uses this function from an innvalid context and it hasn't blown
up. So I suspect that all current users simply use the function from the
proper context. So if nobody objects I would just post the patch for
inclusion. If the compilation breaks we can think of a proper
implementation.
>
> Thanks.
>
> ----------->8---------------
> diff --git a/mm/page_ext.c b/mm/page_ext.c
> index 88ccc044..e3db259 100644
> --- a/mm/page_ext.c
> +++ b/mm/page_ext.c
> @@ -384,6 +384,7 @@ void __init page_ext_init(void)
>
> for_each_node_state(nid, N_MEMORY) {
> unsigned long start_pfn, end_pfn;
> + int page_nid;
>
> start_pfn = node_start_pfn(nid);
> end_pfn = node_end_pfn(nid);
> @@ -405,8 +406,15 @@ void __init page_ext_init(void)
> *
> * Take into account DEFERRED_STRUCT_PAGE_INIT.
> */
> - if (early_pfn_to_nid(pfn) != nid)
> +#ifdef CONFIG_DEFERRED_STRUCT_PAGE_INIT
> + page_nid = early_pfn_to_nid(pfn);
> +#else
> + page_nid = pfn_to_nid(pfn);
> +#endif
I cannot say I would be happy about this ifdefery. Especially when there
is no existing user which would need it.
> +
> + if (page_nid != nid)
> continue;
> +
> if (init_section_page_ext(pfn, nid))
> goto oom;
> }
--
Michal Hocko
SUSE Labs
--
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>
