Re: [PATCHv6 08/17] mm: Make page_zonenum() use head page

[Kiryl Shutsemau <kas@kernel.org>]

On Tue, Feb 10, 2026 at 04:57:55PM +0100, Vlastimil Babka wrote:
> On 2/9/26 12:52, Kiryl Shutsemau wrote:
> > On Thu, Feb 05, 2026 at 02:10:40PM +0100, David Hildenbrand (Arm) wrote:
> >> On 2/2/26 16:56, Kiryl Shutsemau wrote:
> >> > With the upcoming changes to HVO, a single page of tail struct pages
> >> > will be shared across all huge pages of the same order on a node. Since
> >> > huge pages on the same node may belong to different zones, the zone
> >> > information stored in shared tail page flags would be incorrect.
> >> > 
> >> > Always fetch zone information from the head page, which has unique and
> >> > correct zone flags for each compound page.
> >> > 
> >> > Signed-off-by: Kiryl Shutsemau <kas@kernel.org>
> >> > Acked-by: Zi Yan <ziy@nvidia.com>
> >> > ---
> >> >   include/linux/mmzone.h | 1 +
> >> >   1 file changed, 1 insertion(+)
> >> > 
> >> > diff --git a/include/linux/mmzone.h b/include/linux/mmzone.h
> >> > index be8ce40b5638..192143b5cdc0 100644
> >> > --- a/include/linux/mmzone.h
> >> > +++ b/include/linux/mmzone.h
> >> > @@ -1219,6 +1219,7 @@ static inline enum zone_type memdesc_zonenum(memdesc_flags_t flags)
> >> >   static inline enum zone_type page_zonenum(const struct page *page)
> >> >   {
> >> > +	page = compound_head(page);
> >> >   	return memdesc_zonenum(page->flags);
> >> 
> >> We end up calling page_zonenum() without holding a reference.
> >> 
> >> Given that _compound_head() does a READ_ONCE(), this should work even if we
> >> see concurrent page freeing etc.
> >> 
> >> However, this change implies that we now perform a compound page lookup for
> >> every PageHighMem() [meh], page_zone() [quite some users in the buddy,
> >> including for pageblock access and page freeing].
> >> 
> >> That's a nasty compromise for making HVO better? :)
> >> 
> >> We should likely limit that special casing to kernels that really rquire it
> >> (HVO).
> > 
> > I will add compound_info_has_mask() check.
> 
> Not thrilled by this indeed. Would it be a problem to have the shared tail
> pages per node+zone instead of just per node?

I thought it would be overkill. It likely is going to be unused for most
nodes. But sure, move it to per-zone.

-- 
  Kiryl Shutsemau / Kirill A. Shutemov