[PATCH] mm: fix handling zero for large folios with partial zeromap

[Barry Song]

There could be a corner case where the first entry is non-zeromap,
but a subsequent entry is zeromap. In this case, we should not
return false.

Additionally, the iteration of test_bit() is unnecessary and
can be replaced with bitmap operations, which are more efficient.

Since swap_read_folio() can't handle reading a large folio that's
partially zeromap and partially non-zeromap, we've moved the code
to mm/swap.h so that others, like those working on swap-in, can
access it.

Fixes: 0ca0c24e3211 ("mm: store zero pages to be swapped out in a bitmap")
Signed-off-by: Barry Song <v-songbaohua@xxxxxxxx>
---
 mm/page_io.c | 27 ++++-----------------------
 mm/swap.h    | 29 +++++++++++++++++++++++++++++
 2 files changed, 33 insertions(+), 23 deletions(-)

diff --git a/mm/page_io.c b/mm/page_io.c
index 4bc77d1c6bfa..46907c9dd20b 100644
--- a/mm/page_io.c
+++ b/mm/page_io.c
@@ -226,26 +226,6 @@ static void swap_zeromap_folio_clear(struct folio *folio)
 	}
 }
 
-/*
- * Return the index of the first subpage which is not zero-filled
- * according to swap_info_struct->zeromap.
- * If all pages are zero-filled according to zeromap, it will return
- * folio_nr_pages(folio).
- */
-static unsigned int swap_zeromap_folio_test(struct folio *folio)
-{
-	struct swap_info_struct *sis = swp_swap_info(folio->swap);
-	swp_entry_t entry;
-	unsigned int i;
-
-	for (i = 0; i < folio_nr_pages(folio); i++) {
-		entry = page_swap_entry(folio_page(folio, i));
-		if (!test_bit(swp_offset(entry), sis->zeromap))
-			return i;
-	}
-	return i;
-}
-
 /*
  * We may have stale swap cache pages in memory: notice
  * them here and get rid of the unnecessary final write.
@@ -524,9 +504,10 @@ static void sio_read_complete(struct kiocb *iocb, long ret)
 
 static bool swap_read_folio_zeromap(struct folio *folio)
 {
-	unsigned int idx = swap_zeromap_folio_test(folio);
+	unsigned int nr_pages = folio_nr_pages(folio);
+	zeromap_stat_t stat = swap_zeromap_entries_check(folio->swap, nr_pages);
 
-	if (idx == 0)
+	if (stat == SWAP_ZEROMAP_NON)
 		return false;
 
 	/*
@@ -534,7 +515,7 @@ static bool swap_read_folio_zeromap(struct folio *folio)
 	 * currently handled. Return true without marking the folio uptodate so
 	 * that an IO error is emitted (e.g. do_swap_page() will sigbus).
 	 */
-	if (WARN_ON_ONCE(idx < folio_nr_pages(folio)))
+	if (WARN_ON_ONCE(stat == SWAP_ZEROMAP_PARTIAL))
 		return true;
 
 	folio_zero_range(folio, 0, folio_size(folio));
diff --git a/mm/swap.h b/mm/swap.h
index f8711ff82f84..f8e3fa061c1d 100644
--- a/mm/swap.h
+++ b/mm/swap.h
@@ -4,6 +4,12 @@
 
 struct mempolicy;
 
+typedef enum {
+	SWAP_ZEROMAP_NON,
+	SWAP_ZEROMAP_FULL,
+	SWAP_ZEROMAP_PARTIAL
+} zeromap_stat_t;
+
 #ifdef CONFIG_SWAP
 #include <linux/swapops.h> /* for swp_offset */
 #include <linux/blk_types.h> /* for bio_end_io_t */
@@ -80,6 +86,24 @@ static inline unsigned int folio_swap_flags(struct folio *folio)
 {
 	return swp_swap_info(folio->swap)->flags;
 }
+
+/*
+ * Check if nr entries are all zeromap, non-zeromap or partially zeromap
+ */
+static inline zeromap_stat_t swap_zeromap_entries_check(swp_entry_t entry, int nr)
+{
+	struct swap_info_struct *sis = swp_swap_info(entry);
+	unsigned long start = swp_offset(entry);
+	unsigned long end = start + nr;
+
+	if (find_next_bit(sis->zeromap, end, start) == end)
+		return SWAP_ZEROMAP_NON;
+	if (find_next_zero_bit(sis->zeromap, end, start) == end)
+		return SWAP_ZEROMAP_FULL;
+
+	return SWAP_ZEROMAP_PARTIAL;
+}
+
 #else /* CONFIG_SWAP */
 struct swap_iocb;
 static inline void swap_read_folio(struct folio *folio, struct swap_iocb **plug)
@@ -171,6 +195,11 @@ static inline unsigned int folio_swap_flags(struct folio *folio)
 {
 	return 0;
 }
+
+static inline zeromap_stat_t swap_zeromap_entries_check(swp_entry_t entry, int nr)
+{
+	return SWAP_ZEROMAP_NONE;
+}
 #endif /* CONFIG_SWAP */
 
 #endif /* _MM_SWAP_H */
-- 
2.34.1