Demotion will migrate pages across nodes. Previously, only the global
demotion statistics were accounted for. Changed them to per-node
statistics, making it easier to observe where demotion occurs on each
node.
This will help to identify which nodes are under pressure.
This patch also make pgdemote_* behind CONFIG_NUMA_BALANCING, since
demotion is not available for !CONFIG_NUMA_BALANCING
With this patch, here is a sample where node0 node1 are DRAM,
node3 is PMEM:
Global stats:
$ grep demote /proc/vmstat
pgdemote_kswapd 254288
pgdemote_direct 113497
pgdemote_khugepaged 0
Per-node stats:
$ grep demote /sys/devices/system/node/node0/vmstat # demotion source
pgdemote_kswapd 68454
pgdemote_direct 83431
pgdemote_khugepaged 0
$ grep demote /sys/devices/system/node/node1/vmstat # demotion source
pgdemote_kswapd 185834
pgdemote_direct 30066
pgdemote_khugepaged 0
$ grep demote /sys/devices/system/node/node3/vmstat # demotion target
pgdemote_kswapd 0
pgdemote_direct 0
pgdemote_khugepaged 0
Reported-by: kernel test robot <[email protected]> # compling errors
Closes: https://lore.kernel.org/oe-kbuild-all/[email protected]/
Acked-by: "Huang, Ying" <[email protected]>
Signed-off-by: Li Zhijian <[email protected]>
---
V2: split it as a separate patch from previous patch set.
account them to the source node instead destination and add Acked-by # Huang, Ying
---
include/linux/mmzone.h | 4 ++++
include/linux/vm_event_item.h | 3 ---
mm/vmscan.c | 12 ++++++++----
mm/vmstat.c | 6 +++---
4 files changed, 15 insertions(+), 10 deletions(-)
diff --git a/include/linux/mmzone.h b/include/linux/mmzone.h
index 4106fbc5b4b3..ad0309eea850 100644
--- a/include/linux/mmzone.h
+++ b/include/linux/mmzone.h
@@ -206,6 +206,10 @@ enum node_stat_item {
#ifdef CONFIG_NUMA_BALANCING
PGPROMOTE_SUCCESS, /* promote successfully */
PGPROMOTE_CANDIDATE, /* candidate pages to promote */
+ /* PGDEMOTE_*: pages demoted */
+ PGDEMOTE_KSWAPD,
+ PGDEMOTE_DIRECT,
+ PGDEMOTE_KHUGEPAGED,
#endif
NR_VM_NODE_STAT_ITEMS
};
diff --git a/include/linux/vm_event_item.h b/include/linux/vm_event_item.h
index 8abfa1240040..d1b847502f09 100644
--- a/include/linux/vm_event_item.h
+++ b/include/linux/vm_event_item.h
@@ -41,9 +41,6 @@ enum vm_event_item { PGPGIN, PGPGOUT, PSWPIN, PSWPOUT,
PGSTEAL_KSWAPD,
PGSTEAL_DIRECT,
PGSTEAL_KHUGEPAGED,
- PGDEMOTE_KSWAPD,
- PGDEMOTE_DIRECT,
- PGDEMOTE_KHUGEPAGED,
PGSCAN_KSWAPD,
PGSCAN_DIRECT,
PGSCAN_KHUGEPAGED,
diff --git a/mm/vmscan.c b/mm/vmscan.c
index 6f13394b112e..cc70dcefc60a 100644
--- a/mm/vmscan.c
+++ b/mm/vmscan.c
@@ -1110,12 +1110,14 @@ void drop_slab(void)
static int reclaimer_offset(void)
{
+#ifdef CONFIG_NUMA_BALANCING
BUILD_BUG_ON(PGSTEAL_DIRECT - PGSTEAL_KSWAPD !=
PGDEMOTE_DIRECT - PGDEMOTE_KSWAPD);
- BUILD_BUG_ON(PGSTEAL_DIRECT - PGSTEAL_KSWAPD !=
- PGSCAN_DIRECT - PGSCAN_KSWAPD);
BUILD_BUG_ON(PGSTEAL_KHUGEPAGED - PGSTEAL_KSWAPD !=
PGDEMOTE_KHUGEPAGED - PGDEMOTE_KSWAPD);
+#endif
+ BUILD_BUG_ON(PGSTEAL_DIRECT - PGSTEAL_KSWAPD !=
+ PGSCAN_DIRECT - PGSCAN_KSWAPD);
BUILD_BUG_ON(PGSTEAL_KHUGEPAGED - PGSTEAL_KSWAPD !=
PGSCAN_KHUGEPAGED - PGSCAN_KSWAPD);
@@ -1677,8 +1679,10 @@ static unsigned int demote_folio_list(struct list_head *demote_folios,
migrate_pages(demote_folios, alloc_demote_folio, NULL,
(unsigned long)&mtc, MIGRATE_ASYNC, MR_DEMOTION,
&nr_succeeded);
-
- __count_vm_events(PGDEMOTE_KSWAPD + reclaimer_offset(), nr_succeeded);
+#ifdef CONFIG_NUMA_BALANCING
+ mod_node_page_state(pgdat, PGDEMOTE_KSWAPD + reclaimer_offset(),
+ nr_succeeded);
+#endif
return nr_succeeded;
}
diff --git a/mm/vmstat.c b/mm/vmstat.c
index 00e81e99c6ee..f141c48c39e4 100644
--- a/mm/vmstat.c
+++ b/mm/vmstat.c
@@ -1244,6 +1244,9 @@ const char * const vmstat_text[] = {
#ifdef CONFIG_NUMA_BALANCING
"pgpromote_success",
"pgpromote_candidate",
+ "pgdemote_kswapd",
+ "pgdemote_direct",
+ "pgdemote_khugepaged",
#endif
/* enum writeback_stat_item counters */
@@ -1275,9 +1278,6 @@ const char * const vmstat_text[] = {
"pgsteal_kswapd",
"pgsteal_direct",
"pgsteal_khugepaged",
- "pgdemote_kswapd",
- "pgdemote_direct",
- "pgdemote_khugepaged",
"pgscan_kswapd",
"pgscan_direct",
"pgscan_khugepaged",
--
2.29.2
Andrew,
On 03/11/2023 11:14, Li Zhijian wrote:
> Demotion will migrate pages across nodes. Previously, only the global
> demotion statistics were accounted for. Changed them to per-node
> statistics, making it easier to observe where demotion occurs on each
> node.
>
> This will help to identify which nodes are under pressure.
>
> This patch also make pgdemote_* behind CONFIG_NUMA_BALANCING, since
> demotion is not available for !CONFIG_NUMA_BALANCING
I just realized that moving pgdemote_* behind CONFIG_NUMA_BALANCING was wrong.
Demotion works well without CONFIG_NUMA_BALANCING.
Since this patch was already in the mm-stable branch, is it possible to
replace this patch from mm-stable or post a fixup for this
Thanks
Zhijian
>
> With this patch, here is a sample where node0 node1 are DRAM,
> node3 is PMEM:
> Global stats:
> $ grep demote /proc/vmstat
> pgdemote_kswapd 254288
> pgdemote_direct 113497
> pgdemote_khugepaged 0
>
> Per-node stats:
> $ grep demote /sys/devices/system/node/node0/vmstat # demotion source
> pgdemote_kswapd 68454
> pgdemote_direct 83431
> pgdemote_khugepaged 0
> $ grep demote /sys/devices/system/node/node1/vmstat # demotion source
> pgdemote_kswapd 185834
> pgdemote_direct 30066
> pgdemote_khugepaged 0
> $ grep demote /sys/devices/system/node/node3/vmstat # demotion target
> pgdemote_kswapd 0
> pgdemote_direct 0
> pgdemote_khugepaged 0
>
> Reported-by: kernel test robot <[email protected]> # compling errors
> Closes: https://lore.kernel.org/oe-kbuild-all/[email protected]/
> Acked-by: "Huang, Ying" <[email protected]>
> Signed-off-by: Li Zhijian <[email protected]>
>
> ---
> V2: split it as a separate patch from previous patch set.
> account them to the source node instead destination and add Acked-by # Huang, Ying
> ---
> include/linux/mmzone.h | 4 ++++
> include/linux/vm_event_item.h | 3 ---
> mm/vmscan.c | 12 ++++++++----
> mm/vmstat.c | 6 +++---
> 4 files changed, 15 insertions(+), 10 deletions(-)
>
> diff --git a/include/linux/mmzone.h b/include/linux/mmzone.h
> index 4106fbc5b4b3..ad0309eea850 100644
> --- a/include/linux/mmzone.h
> +++ b/include/linux/mmzone.h
> @@ -206,6 +206,10 @@ enum node_stat_item {
> #ifdef CONFIG_NUMA_BALANCING
> PGPROMOTE_SUCCESS, /* promote successfully */
> PGPROMOTE_CANDIDATE, /* candidate pages to promote */
> + /* PGDEMOTE_*: pages demoted */
> + PGDEMOTE_KSWAPD,
> + PGDEMOTE_DIRECT,
> + PGDEMOTE_KHUGEPAGED,
> #endif
> NR_VM_NODE_STAT_ITEMS
> };
> diff --git a/include/linux/vm_event_item.h b/include/linux/vm_event_item.h
> index 8abfa1240040..d1b847502f09 100644
> --- a/include/linux/vm_event_item.h
> +++ b/include/linux/vm_event_item.h
> @@ -41,9 +41,6 @@ enum vm_event_item { PGPGIN, PGPGOUT, PSWPIN, PSWPOUT,
> PGSTEAL_KSWAPD,
> PGSTEAL_DIRECT,
> PGSTEAL_KHUGEPAGED,
> - PGDEMOTE_KSWAPD,
> - PGDEMOTE_DIRECT,
> - PGDEMOTE_KHUGEPAGED,
> PGSCAN_KSWAPD,
> PGSCAN_DIRECT,
> PGSCAN_KHUGEPAGED,
> diff --git a/mm/vmscan.c b/mm/vmscan.c
> index 6f13394b112e..cc70dcefc60a 100644
> --- a/mm/vmscan.c
> +++ b/mm/vmscan.c
> @@ -1110,12 +1110,14 @@ void drop_slab(void)
>
> static int reclaimer_offset(void)
> {
> +#ifdef CONFIG_NUMA_BALANCING
> BUILD_BUG_ON(PGSTEAL_DIRECT - PGSTEAL_KSWAPD !=
> PGDEMOTE_DIRECT - PGDEMOTE_KSWAPD);
> - BUILD_BUG_ON(PGSTEAL_DIRECT - PGSTEAL_KSWAPD !=
> - PGSCAN_DIRECT - PGSCAN_KSWAPD);
> BUILD_BUG_ON(PGSTEAL_KHUGEPAGED - PGSTEAL_KSWAPD !=
> PGDEMOTE_KHUGEPAGED - PGDEMOTE_KSWAPD);
> +#endif
> + BUILD_BUG_ON(PGSTEAL_DIRECT - PGSTEAL_KSWAPD !=
> + PGSCAN_DIRECT - PGSCAN_KSWAPD);
> BUILD_BUG_ON(PGSTEAL_KHUGEPAGED - PGSTEAL_KSWAPD !=
> PGSCAN_KHUGEPAGED - PGSCAN_KSWAPD);
>
> @@ -1677,8 +1679,10 @@ static unsigned int demote_folio_list(struct list_head *demote_folios,
> migrate_pages(demote_folios, alloc_demote_folio, NULL,
> (unsigned long)&mtc, MIGRATE_ASYNC, MR_DEMOTION,
> &nr_succeeded);
> -
> - __count_vm_events(PGDEMOTE_KSWAPD + reclaimer_offset(), nr_succeeded);
> +#ifdef CONFIG_NUMA_BALANCING
> + mod_node_page_state(pgdat, PGDEMOTE_KSWAPD + reclaimer_offset(),
> + nr_succeeded);
> +#endif
>
> return nr_succeeded;
> }
> diff --git a/mm/vmstat.c b/mm/vmstat.c
> index 00e81e99c6ee..f141c48c39e4 100644
> --- a/mm/vmstat.c
> +++ b/mm/vmstat.c
> @@ -1244,6 +1244,9 @@ const char * const vmstat_text[] = {
> #ifdef CONFIG_NUMA_BALANCING
> "pgpromote_success",
> "pgpromote_candidate",
> + "pgdemote_kswapd",
> + "pgdemote_direct",
> + "pgdemote_khugepaged",
> #endif
>
> /* enum writeback_stat_item counters */
> @@ -1275,9 +1278,6 @@ const char * const vmstat_text[] = {
> "pgsteal_kswapd",
> "pgsteal_direct",
> "pgsteal_khugepaged",
> - "pgdemote_kswapd",
> - "pgdemote_direct",
> - "pgdemote_khugepaged",
> "pgscan_kswapd",
> "pgscan_direct",
> "pgscan_khugepaged",
On Thu, 28 Dec 2023 03:52:25 +0000 "Zhijian Li (Fujitsu)" <[email protected]> wrote:
> > This patch also make pgdemote_* behind CONFIG_NUMA_BALANCING, since
> > demotion is not available for !CONFIG_NUMA_BALANCING
>
> I just realized that moving pgdemote_* behind CONFIG_NUMA_BALANCING was wrong.
> Demotion works well without CONFIG_NUMA_BALANCING.
>
> Since this patch was already in the mm-stable branch, is it possible to
> replace this patch from mm-stable or post a fixup for this
A fixup would be preferred, please. With the appropriate Fixes: