For unpopulated zones, the pagesets point to the common
boot_pageset which can have non-zero vm_numa_stat counts.
Because of this memory-less nodes end up having non-zero
NUMA statistics. This can be observed on any architecture
that supports memory-less NUMA nodes.
E.g.
$ numactl -H
available: 2 nodes (0-1)
node 0 cpus: 0 1 2 3
node 0 size: 0 MB
node 0 free: 0 MB
node 1 cpus: 4 5 6 7
node 1 size: 8131 MB
node 1 free: 6980 MB
node distances:
node 0 1
0: 10 40
1: 40 10
$ numastat
node0 node1
numa_hit 108 56495
numa_miss 0 0
numa_foreign 0 0
interleave_hit 0 4537
local_node 108 31547
other_node 0 24948
Hence, return zero explicitly for all the stats of an
unpopulated zone.
Signed-off-by: Sandipan Das <sandipan@xxxxxxxxxxxxx>
---
include/linux/vmstat.h | 15 +++++++++++++++
1 file changed, 15 insertions(+)
diff --git a/include/linux/vmstat.h b/include/linux/vmstat.h
index 292485f3d24d..55a68b379a2c 100644
--- a/include/linux/vmstat.h
+++ b/include/linux/vmstat.h
@@ -159,6 +159,21 @@ static inline unsigned long zone_numa_state_snapshot(struct zone *zone,
long x = atomic_long_read(&zone->vm_numa_stat[item]);
int cpu;
+ /*
+ * Initially, the pageset of all zones are set to point to the
+ * boot_pageset. The real pagesets are allocated later but only
+ * for the populated zones. Unpopulated zones still continue
+ * using the boot_pageset.
+ *
+ * Before the real pagesets are allocated, the boot_pageset's
+ * vm_numa_stat counters can get incremented. This affects the
+ * unpopulated zones which end up with non-zero stats despite
+ * having no memory associated with them. For such cases,
+ * return zero explicitly.
+ */
+ if (!populated_zone(zone))
+ return 0;
+
for_each_online_cpu(cpu)
x += per_cpu_ptr(zone->pageset, cpu)->vm_numa_stat_diff[item];
--
2.17.1
