[PATCH RFC v2 09/15] mm/madvise: adapt for LRU refs based operations in MGLRU

[Kairui Song via B4 Relay]

From: Kairui Song <kasong@xxxxxxxxxxx>

For the active/inactive LRU, madvise clears PG_referenced so that one more
access is not enough to reactivate a folio, and keeps PG_workingset on a
folio demoted out of the active list so its refault is still accounted as
a workingset refault (PSI).

MGLRU keeps that history in the folio's refs count instead, which now
lives in PG_referenced, PG_workingset and LRU_REFS_MASK, so writing to
those bits directly corrupts the count. The two hints want different
things:

- MADV_COLD resets the count in folio_deactivate(), which also moves the
  folio to the oldest generation.
- MADV_PAGEOUT isolates the folio and hands it to reclaim_pages(), which
  ignores the count, so keep it and let the eviction shadow record the
  folio's real hotness.

PG_young is idle-page-tracking state, not a reclaim input, so it is still
cleared for both LRUs.

Signed-off-by: Kairui Song <kasong@xxxxxxxxxxx>
---
 mm/madvise.c | 48 ++++++++++++++++++++++++++++++++++--------------
 1 file changed, 34 insertions(+), 14 deletions(-)

diff --git a/mm/madvise.c b/mm/madvise.c
index f75a9d139980..cdd1bebf2485 100644
--- a/mm/madvise.c
+++ b/mm/madvise.c
@@ -361,6 +361,38 @@ static inline int madvise_folio_pte_batch(unsigned long addr, unsigned long end,
 				     FPB_MERGE_YOUNG_DIRTY);
 }
 
+/*
+ * Drop the recency information madvise() is expected to drop before
+ * deactivating or reclaiming a folio.
+ */
+static void madvise_cold_or_pageout_prep_folio(struct folio *folio)
+{
+	/*
+	 * VM couldn't reclaim the folio unless we clear PG_young.
+	 * As a side effect, it makes confuse idle-page tracking
+	 * because they will miss recent referenced history.
+	 */
+	folio_test_clear_young(folio);
+
+	/*
+	 * MGLRU keeps this history in the refs count. MADV_COLD resets
+	 * it in folio_deactivate(), MADV_PAGEOUT hands the folio to
+	 * reclaim_pages(), which ignores the count, so keep it, and let
+	 * the eviction shadow record the folio's real hotness.
+	 */
+	if (lru_gen_enabled())
+		return;
+
+	/*
+	 * For the active/inactive LRU, a folio demoted out of the active
+	 * list should have PG_workingset so its refault is still accounted
+	 * as a workingset refault.
+	 */
+	folio_clear_referenced(folio);
+	if (folio_test_active(folio))
+		folio_set_workingset(folio);
+}
+
 static int madvise_cold_or_pageout_pte_range(pmd_t *pmd,
 				unsigned long addr, unsigned long end,
 				struct mm_walk *walk)
@@ -438,10 +470,7 @@ static int madvise_cold_or_pageout_pte_range(pmd_t *pmd,
 			tlb_remove_pmd_tlb_entry(tlb, pmd, addr);
 		}
 
-		folio_clear_referenced(folio);
-		folio_test_clear_young(folio);
-		if (folio_test_active(folio))
-			folio_set_workingset(folio);
+		madvise_cold_or_pageout_prep_folio(folio);
 		if (pageout) {
 			if (folio_isolate_lru(folio)) {
 				if (folio_test_unevictable(folio))
@@ -547,16 +576,7 @@ static int madvise_cold_or_pageout_pte_range(pmd_t *pmd,
 			tlb_remove_tlb_entries(tlb, pte, nr, addr);
 		}
 
-		/*
-		 * We are deactivating a folio for accelerating reclaiming.
-		 * VM couldn't reclaim the folio unless we clear PG_young.
-		 * As a side effect, it makes confuse idle-page tracking
-		 * because they will miss recent referenced history.
-		 */
-		folio_clear_referenced(folio);
-		folio_test_clear_young(folio);
-		if (folio_test_active(folio))
-			folio_set_workingset(folio);
+		madvise_cold_or_pageout_prep_folio(folio);
 		if (pageout) {
 			if (folio_isolate_lru(folio)) {
 				if (folio_test_unevictable(folio))

-- 
2.55.0