[PATCH v2] vmstat: call fold_vm_zone_numa_events() before show per zone NUMA event

[PATCH v2] vmstat: call fold_vm_zone_numa_events() before show per zone NUMA event

[mengensun88]

From: MengEn Sun <mengensun@tencent.com>

Since 5.14-rc1, NUMA events will only be folded from per-CPU
statistics to per zone and global statistics when the user
actually needs it.

Currently, the kernel has performs the fold operation when reading
/proc/vmstat, but does not perform the fold operation in
/proc/zoneinfo. This can lead to inaccuracies in the following
statistics in zoneinfo:
- numa_hit
- numa_miss
- numa_foreign
- numa_interleave
- numa_local
- numa_other

Therefore, before printing per-zone vm_numa_event when reading
/proc/zoneinfo, we should also perform the fold operation.

Fixes: f19298b9516c ("mm/vmstat: convert NUMA statistics to basic NUMA counters")
Reviewed-by: JinLiang Zheng <alexjlzheng@tencent.com>
Signed-off-by: MengEn Sun <mengensun@tencent.com>
---
changelog:
v1: https://lore.kernel.org/linux-mm/20241030163527.cc9075a4aeb60448873aea73@linux-foundation.org/
v2: made some modifications to the commit message: 
    - Made the commit information more detailed
    - Added the range of versions that need to be fixed.
---
 mm/vmstat.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/mm/vmstat.c b/mm/vmstat.c
index b5a4cea..2770800 100644
--- a/mm/vmstat.c
+++ b/mm/vmstat.c
@@ -1778,6 +1778,7 @@ static void zoneinfo_show_print(struct seq_file *m, pg_data_t *pgdat,
 			   zone_page_state(zone, i));
 
 #ifdef CONFIG_NUMA
+	fold_vm_zone_numa_events(zone);
 	for (i = 0; i < NR_VM_NUMA_EVENT_ITEMS; i++)
 		seq_printf(m, "\n      %-12s %lu", numa_stat_name(i),
 			   zone_numa_event_state(zone, i));
-- 
1.8.3.1

Re: [PATCH v2] vmstat: call fold_vm_zone_numa_events() before show per zone NUMA event

[Andrew Morton]

On Fri,  1 Nov 2024 12:06:38 +0800 mengensun88@gmail.com wrote:

> Since 5.14-rc1, NUMA events will only be folded from per-CPU
> statistics to per zone and global statistics when the user
> actually needs it.
> 
> Currently, the kernel has performs the fold operation when reading
> /proc/vmstat, but does not perform the fold operation in
> /proc/zoneinfo. This can lead to inaccuracies in the following
> statistics in zoneinfo:
> - numa_hit
> - numa_miss
> - numa_foreign
> - numa_interleave
> - numa_local
> - numa_other

Maybe we should backport this.  Please tell us more about these
inaccuracies.  How large are they and what is the impact upon our
users?