[PATCH v3 05/11] mm/memory_hotplug: Optimize zone shrinking code when checking for holes

[David Hildenbrand <david@redhat.com>]

... and clarify why this is needed at all right now. It all boils down
to false positives. We will try to remove the false positives for
!ZONE_DEVICE memory, soon, however, for ZONE_DEVICE memory we won't be
able to easily get rid of false positives.

Don't only detect "all holes" but try to shrink using the existing
functions we have.

Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Oscar Salvador <osalvador@suse.de>
Cc: David Hildenbrand <david@redhat.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Pavel Tatashin <pasha.tatashin@soleen.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Wei Yang <richardw.yang@linux.intel.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
---
 mm/memory_hotplug.c | 45 +++++++++++++++++++++++----------------------
 1 file changed, 23 insertions(+), 22 deletions(-)

diff --git a/mm/memory_hotplug.c b/mm/memory_hotplug.c
index d3c34bbeb36d..663853bf97ed 100644
--- a/mm/memory_hotplug.c
+++ b/mm/memory_hotplug.c
@@ -411,32 +411,33 @@ static void shrink_zone_span(struct zone *zone, unsigned long start_pfn,
 		}
 	}
 
-	/*
-	 * The section is not biggest or smallest mem_section in the zone, it
-	 * only creates a hole in the zone. So in this case, we need not
-	 * change the zone. But perhaps, the zone has only hole data. Thus
-	 * it check the zone has only hole or not.
-	 */
-	for (pfn = zone->zone_start_pfn;
-	     pfn < zone_end_pfn(zone); pfn += PAGES_PER_SUBSECTION) {
-		if (unlikely(!pfn_valid(pfn)))
-			continue;
-
-		if (page_zone(pfn_to_page(pfn)) != zone)
-			continue;
-
-		/* Skip range to be removed */
-		if (pfn >= start_pfn && pfn < end_pfn)
-			continue;
-
-		/* If we find valid section, we have nothing to do */
+	if (!zone->spanned_pages) {
 		zone_span_writeunlock(zone);
 		return;
 	}
 
-	/* The zone has no valid section */
-	zone->zone_start_pfn = 0;
-	zone->spanned_pages = 0;
+	/*
+	 * Due to false positives in previous skrink attempts, it can happen
+	 * that we can shrink the zones further (possibly to zero). Once we
+	 * can reliably detect which PFNs actually belong to a zone
+	 * (especially for ZONE_DEVICE memory where we don't have online
+	 * sections), this can go.
+	 */
+	pfn = find_smallest_section_pfn(nid, zone, zone->zone_start_pfn,
+					zone_end_pfn(zone));
+	if (pfn) {
+		zone->spanned_pages = zone_end_pfn(zone) - pfn;
+		zone->zone_start_pfn = pfn;
+
+		pfn = find_biggest_section_pfn(nid, zone, zone->zone_start_pfn,
+					       zone_end_pfn(zone));
+		if (pfn)
+			zone->spanned_pages = pfn - zone->zone_start_pfn + 1;
+	}
+	if (!pfn) {
+		zone->zone_start_pfn = 0;
+		zone->spanned_pages = 0;
+	}
 	zone_span_writeunlock(zone);
 }
 
-- 
2.21.0