RE: [PATCH v7 1/2] mm/memory_hotplug: make shrink_zone_span() more robust

> -----Original Message-----
> From: Liu, Yuan1 <[email protected]>
> Sent: Tuesday, August 18, 2026 4:57 PM
> To: David Hildenbrand <[email protected]>; Oscar Salvador
> <[email protected]>; Mike Rapoport <[email protected]>; Wei Yang
> <[email protected]>
> Cc: [email protected]; Zou, Nanhai <[email protected]>; Chen Zhang
> <[email protected]>; Liu, Yuan1 <[email protected]>; Zeng, Jason
> <[email protected]>; Chen, Yu C <[email protected]>; Deng, Pan
> <[email protected]>; Li, Tianyou <[email protected]>; linux-
> [email protected]
> Subject: [PATCH v7 1/2] mm/memory_hotplug: make shrink_zone_span() more
> robust
> 
> From: "David Hildenbrand (Arm)" <[email protected]>
> 
> Let's make shrink_zone_span() more robust by checking in
> find_smallest_section_pfn() / find_biggest_section_pfn() that the
> start and end PFNs of the subsection are within the zone.
> 
> While at it, clean up the function by factoring the core check out
> into subsection_overlaps_zone().
> 
> There likely is no need to check the nid first. We require
> SPARSEMEM_VMEMMAP_ENABLE, where pfn_to_page() is cheap, and
> pfn_to_nid() on CONFIG_NUMA would call pfn_to_page() either way.
> So let's just drop that for now.
> 
> Signed-off-by: David Hildenbrand (Arm) <[email protected]>
> Tested-by: Yuan Liu <[email protected]>
> Signed-off-by: Yuan Liu <[email protected]>
> ---
>  mm/memory_hotplug.c | 59 ++++++++++++++++++---------------------------
>  1 file changed, 24 insertions(+), 35 deletions(-)
> 
> diff --git a/mm/memory_hotplug.c b/mm/memory_hotplug.c
> index 7ac19fab2263..cd82e79f0782 100644
> --- a/mm/memory_hotplug.c
> +++ b/mm/memory_hotplug.c
[...]

>  /* find the biggest valid pfn in the range [start_pfn, end_pfn). */
> -static unsigned long find_biggest_section_pfn(int nid, struct zone *zone,
> -				    unsigned long start_pfn,
> -				    unsigned long end_pfn)
> +static unsigned long find_biggest_section_pfn(struct zone *zone,
> +		unsigned long start_pfn, unsigned long end_pfn)
>  {
> -	unsigned long pfn;
> -
> -	/* pfn is the end pfn of a memory section. */
> -	pfn = end_pfn - 1;
> -	for (; pfn >= start_pfn; pfn -= PAGES_PER_SUBSECTION) {
> -		if (unlikely(!pfn_to_online_page(pfn)))
> -			continue;
> -
> -		if (unlikely(pfn_to_nid(pfn) != nid))
> -			continue;
> -
> -		if (zone != page_zone(pfn_to_page(pfn)))
> -			continue;
> -
> -		return pfn;
> +	for (; end_pfn >= start_pfn; end_pfn -= PAGES_PER_SUBSECTION) {
> +		if (subsection_overlaps_zone(end_pfn - 1, zone))
> +			return end_pfn - 1;
>  	}

Hi David

Sashiko[1] pointed out that the loop can evaluate an out-of-range PFN
and underflow when start_pfn is 0. I'll change the condition to
end_pfn > start_pfn to address this.

[1] https://sashiko.dev/#/patchset/20260818085702.3395529-1-yuan1.liu%40intel.com

Best Regards,
Liu, Yuan

> -
>  	return 0;
>  }