[PATCH] mm: vmscan: rework lru_shrink and write_folio tracepoints

[PATCH] mm: vmscan: rework lru_shrink and write_folio tracepoints

[qiwu.chen]

Currently, reclaim_flags always contains RECLAIM_WB_ASYNC in lru_shrink
tracepoints since commit 41ac1999c3e35 ("mm: vmscan: do not stall on
writeback during memory compaction"), which is useless for debugging
memory pressure issues. Other RECLAIM_WB_* flags are not used anywhere
else, so they can be directly removed.
This patch reworks the lru_shrink and write_folio tracepoints for better
correlation and analysis:
 - traces each folio lru type instead of reclaim_flags.
 - traces each lru_shrink with reason.

Fixes: 41ac1999c3e35 ("mm: vmscan: do not stall on writeback during memory compaction")
Signed-off-by: qiwu.chen <qiwu.chen@transsion.com>
---
 include/trace/events/vmscan.h | 65 +++++++++++++++--------------------
 mm/vmscan.c                   |  9 ++---
 2 files changed, 32 insertions(+), 42 deletions(-)

diff --git a/include/trace/events/vmscan.h b/include/trace/events/vmscan.h
index 4445a8d9218d..d0a7fcd265e2 100644
--- a/include/trace/events/vmscan.h
+++ b/include/trace/events/vmscan.h
@@ -11,22 +11,6 @@
 #include <linux/memcontrol.h>
 #include <trace/events/mmflags.h>
 
-#define RECLAIM_WB_ANON		0x0001u
-#define RECLAIM_WB_FILE		0x0002u
-#define RECLAIM_WB_MIXED	0x0010u
-#define RECLAIM_WB_SYNC		0x0004u /* Unused, all reclaim async */
-#define RECLAIM_WB_ASYNC	0x0008u
-#define RECLAIM_WB_LRU		(RECLAIM_WB_ANON|RECLAIM_WB_FILE)
-
-#define show_reclaim_flags(flags)				\
-	(flags) ? __print_flags(flags, "|",			\
-		{RECLAIM_WB_ANON,	"RECLAIM_WB_ANON"},	\
-		{RECLAIM_WB_FILE,	"RECLAIM_WB_FILE"},	\
-		{RECLAIM_WB_MIXED,	"RECLAIM_WB_MIXED"},	\
-		{RECLAIM_WB_SYNC,	"RECLAIM_WB_SYNC"},	\
-		{RECLAIM_WB_ASYNC,	"RECLAIM_WB_ASYNC"}	\
-		) : "RECLAIM_WB_NONE"
-
 #define _VMSCAN_THROTTLE_WRITEBACK	(1 << VMSCAN_THROTTLE_WRITEBACK)
 #define _VMSCAN_THROTTLE_ISOLATED	(1 << VMSCAN_THROTTLE_ISOLATED)
 #define _VMSCAN_THROTTLE_NOPROGRESS	(1 << VMSCAN_THROTTLE_NOPROGRESS)
@@ -51,10 +35,11 @@ TRACE_DEFINE_ENUM(KSWAPD_CLEAR_HOPELESS_PCP);
 	{KSWAPD_CLEAR_HOPELESS_PCP,	"PCP"},		\
 	{KSWAPD_CLEAR_HOPELESS_OTHER,	"OTHER"}
 
-#define trace_reclaim_flags(file) ( \
-	(file ? RECLAIM_WB_FILE : RECLAIM_WB_ANON) | \
-	(RECLAIM_WB_ASYNC) \
-	)
+#define trace_reclaim_reason_ops		\
+	{PGSTEAL_KSWAPD,	"KSWAPD"},	\
+	{PGSTEAL_DIRECT,	"DIRECT"},	\
+	{PGSTEAL_KHUGEPAGED,	"KHUGEPAGED"}, \
+	{PGSTEAL_PROACTIVE,	"PROACTIVE"}
 
 TRACE_EVENT(mm_vmscan_kswapd_sleep,
 
@@ -362,19 +347,17 @@ TRACE_EVENT(mm_vmscan_write_folio,
 
 	TP_STRUCT__entry(
 		__field(unsigned long, pfn)
-		__field(int, reclaim_flags)
+		__field(int, lru)
 	),
 
 	TP_fast_assign(
 		__entry->pfn = folio_pfn(folio);
-		__entry->reclaim_flags = trace_reclaim_flags(
-						folio_is_file_lru(folio));
+		__entry->lru = folio_lru_list(folio);
 	),
 
-	TP_printk("page=%p pfn=0x%lx flags=%s",
+	TP_printk("page=%p lru=%s",
 		pfn_to_page(__entry->pfn),
-		__entry->pfn,
-		show_reclaim_flags(__entry->reclaim_flags))
+		__print_symbolic(__entry->lru, LRU_NAMES))
 );
 
 TRACE_EVENT(mm_vmscan_reclaim_pages,
@@ -426,9 +409,9 @@ TRACE_EVENT(mm_vmscan_lru_shrink_inactive,
 
 	TP_PROTO(int nid,
 		unsigned long nr_scanned, unsigned long nr_reclaimed,
-		struct reclaim_stat *stat, int priority, int file),
+		struct reclaim_stat *stat, int priority, int lru, int reason),
 
-	TP_ARGS(nid, nr_scanned, nr_reclaimed, stat, priority, file),
+	TP_ARGS(nid, nr_scanned, nr_reclaimed, stat, priority, lru, reason),
 
 	TP_STRUCT__entry(
 		__field(int, nid)
@@ -443,7 +426,8 @@ TRACE_EVENT(mm_vmscan_lru_shrink_inactive,
 		__field(unsigned long, nr_ref_keep)
 		__field(unsigned long, nr_unmap_fail)
 		__field(int, priority)
-		__field(int, reclaim_flags)
+		__field(int, lru)
+		__field(int, reason)
 	),
 
 	TP_fast_assign(
@@ -459,10 +443,11 @@ TRACE_EVENT(mm_vmscan_lru_shrink_inactive,
 		__entry->nr_ref_keep = stat->nr_ref_keep;
 		__entry->nr_unmap_fail = stat->nr_unmap_fail;
 		__entry->priority = priority;
-		__entry->reclaim_flags = trace_reclaim_flags(file);
+		__entry->lru = lru;
+		__entry->reason = reason;
 	),
 
-	TP_printk("nid=%d nr_scanned=%ld nr_reclaimed=%ld nr_dirty=%ld nr_writeback=%ld nr_congested=%ld nr_immediate=%ld nr_activate_anon=%d nr_activate_file=%d nr_ref_keep=%ld nr_unmap_fail=%ld priority=%d flags=%s",
+	TP_printk("nid=%d nr_scanned=%ld nr_reclaimed=%ld nr_dirty=%ld nr_writeback=%ld nr_congested=%ld nr_immediate=%ld nr_activate_anon=%d nr_activate_file=%d nr_ref_keep=%ld nr_unmap_fail=%ld priority=%d lru=%s reason=%s",
 		__entry->nid,
 		__entry->nr_scanned, __entry->nr_reclaimed,
 		__entry->nr_dirty, __entry->nr_writeback,
@@ -470,16 +455,17 @@ TRACE_EVENT(mm_vmscan_lru_shrink_inactive,
 		__entry->nr_activate0, __entry->nr_activate1,
 		__entry->nr_ref_keep, __entry->nr_unmap_fail,
 		__entry->priority,
-		show_reclaim_flags(__entry->reclaim_flags))
+		__print_symbolic(__entry->lru, LRU_NAMES),
+		__print_symbolic(__entry->reason, trace_reclaim_reason_ops))
 );
 
 TRACE_EVENT(mm_vmscan_lru_shrink_active,
 
 	TP_PROTO(int nid, unsigned long nr_taken,
 		unsigned long nr_active, unsigned long nr_deactivated,
-		unsigned long nr_referenced, int priority, int file),
+		unsigned long nr_referenced, int priority, int lru, int reason),
 
-	TP_ARGS(nid, nr_taken, nr_active, nr_deactivated, nr_referenced, priority, file),
+	TP_ARGS(nid, nr_taken, nr_active, nr_deactivated, nr_referenced, priority, lru, reason),
 
 	TP_STRUCT__entry(
 		__field(int, nid)
@@ -488,7 +474,8 @@ TRACE_EVENT(mm_vmscan_lru_shrink_active,
 		__field(unsigned long, nr_deactivated)
 		__field(unsigned long, nr_referenced)
 		__field(int, priority)
-		__field(int, reclaim_flags)
+		__field(int, lru)
+		__field(int, reason)
 	),
 
 	TP_fast_assign(
@@ -498,15 +485,17 @@ TRACE_EVENT(mm_vmscan_lru_shrink_active,
 		__entry->nr_deactivated = nr_deactivated;
 		__entry->nr_referenced = nr_referenced;
 		__entry->priority = priority;
-		__entry->reclaim_flags = trace_reclaim_flags(file);
+		__entry->lru = lru;
+		__entry->reason = reason;
 	),
 
-	TP_printk("nid=%d nr_taken=%ld nr_active=%ld nr_deactivated=%ld nr_referenced=%ld priority=%d flags=%s",
+	TP_printk("nid=%d nr_taken=%ld nr_active=%ld nr_deactivated=%ld nr_referenced=%ld priority=%d lru=%s reason=%s",
 		__entry->nid,
 		__entry->nr_taken,
 		__entry->nr_active, __entry->nr_deactivated, __entry->nr_referenced,
 		__entry->priority,
-		show_reclaim_flags(__entry->reclaim_flags))
+		__print_symbolic(__entry->lru, LRU_NAMES),
+		__print_symbolic(__entry->reason, trace_reclaim_reason_ops))
 );
 
 TRACE_EVENT(mm_vmscan_node_reclaim_begin,
diff --git a/mm/vmscan.c b/mm/vmscan.c
index bd1b1aa12581..4ee84db91635 100644
--- a/mm/vmscan.c
+++ b/mm/vmscan.c
@@ -2044,7 +2044,7 @@ static unsigned long shrink_inactive_list(unsigned long nr_to_scan,
 		sc->nr.file_taken += nr_taken;
 
 	trace_mm_vmscan_lru_shrink_inactive(pgdat->node_id,
-			nr_scanned, nr_reclaimed, &stat, sc->priority, file);
+			nr_scanned, nr_reclaimed, &stat, sc->priority, lru, item);
 	return nr_reclaimed;
 }
 
@@ -2151,7 +2151,8 @@ static void shrink_active_list(unsigned long nr_to_scan,
 	lruvec_lock_irq(lruvec);
 	lru_note_cost_unlock_irq(lruvec, file, 0, nr_rotated);
 	trace_mm_vmscan_lru_shrink_active(pgdat->node_id, nr_taken, nr_activate,
-			nr_deactivate, nr_rotated, sc->priority, file);
+			nr_deactivate, nr_rotated, sc->priority, lru,
+			PGSTEAL_KSWAPD + reclaimer_offset(sc));
 }
 
 static unsigned int reclaim_folio_list(struct list_head *folio_list,
@@ -4854,9 +4855,10 @@ static int evict_folios(unsigned long nr_to_scan, struct lruvec *lruvec,
 	reclaimed = shrink_folio_list(&list, pgdat, sc, &stat, false, memcg);
 	sc->nr.unqueued_dirty += stat.nr_unqueued_dirty;
 	sc->nr_reclaimed += reclaimed;
+	item = PGSTEAL_KSWAPD + reclaimer_offset(sc);
 	trace_mm_vmscan_lru_shrink_inactive(pgdat->node_id,
 			scanned, reclaimed, &stat, sc->priority,
-			type ? LRU_INACTIVE_FILE : LRU_INACTIVE_ANON);
+			type ? LRU_INACTIVE_FILE : LRU_INACTIVE_ANON, item);
 
 	list_for_each_entry_safe_reverse(folio, next, &list, lru) {
 		DEFINE_MIN_SEQ(lruvec);
@@ -4892,7 +4894,6 @@ static int evict_folios(unsigned long nr_to_scan, struct lruvec *lruvec,
 	mod_lruvec_state(lruvec, PGDEMOTE_KSWAPD + reclaimer_offset(sc),
 					stat.nr_demoted);
 
-	item = PGSTEAL_KSWAPD + reclaimer_offset(sc);
 	mod_lruvec_state(lruvec, item, reclaimed);
 	mod_lruvec_state(lruvec, PGSTEAL_ANON + type, reclaimed);
 
-- 
2.25.1

Re: [PATCH] mm: vmscan: rework lru_shrink and write_folio tracepoints

[Matthew Wilcox]

On Sun, May 03, 2026 at 04:57:05PM +0800, qiwu.chen wrote:
> Currently, reclaim_flags always contains RECLAIM_WB_ASYNC in lru_shrink
> tracepoints since commit 41ac1999c3e35 ("mm: vmscan: do not stall on
> writeback during memory compaction"), which is useless for debugging
> memory pressure issues. Other RECLAIM_WB_* flags are not used anywhere
> else, so they can be directly removed.
> This patch reworks the lru_shrink and write_folio tracepoints for better
> correlation and analysis:
>  - traces each folio lru type instead of reclaim_flags.
>  - traces each lru_shrink with reason.

You also removed the printing of the folio's PFN.  Was this deliberate?
If so, it merits a mention in the commit description.

Also if you are going to do this (and I suspect we should do this!)
we don't need to do the folio -> pfn -> page conversion.  We can just
store the folio pointer and print out the folio pointer.

Re: [PATCH] mm: vmscan: rework lru_shrink and write_folio tracepoints

[chenqiwu]

On Sun, May 03, 2026 at 01:35:21PM +0100, Matthew Wilcox wrote:
> 
> You also removed the printing of the folio's PFN.  Was this deliberate?
> If so, it merits a mention in the commit description.
> 
> Also if you are going to do this (and I suspect we should do this!)
> we don't need to do the folio -> pfn -> page conversion.  We can just
> store the folio pointer and print out the folio pointer.
>
Yes, I think printing out the PFN is indeed unnecessary, but miss to
merit a mention in commit messsage. I will remove the unnecessary
conversion and merit a mention in patch v2.