[RESEND PATCH] mm/page_table_check: add explicit pmd_none check in pte_clear_range

From: Ye Liu <[email protected]>

In __page_table_check_pte_clear_range(), the condition to determine
whether to iterate over PTEs only checked pmd_bad() and pmd_leaf().
This relies on the implicit assumption that pmd_none() is always a
subset of pmd_bad() on all architectures supporting PAGE_TABLE_CHECK.

While this assumption currently holds for x86_64, arm64, s390, riscv,
and powerpc, it is an architecture-dependent behavior that may not
hold for future architectures. Add an explicit pmd_none() check to
make the intent clear and avoid calling pte_offset_map() on an empty
PMD, which could lead to undefined behavior.

Signed-off-by: Ye Liu <[email protected]>
---
 mm/page_table_check.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/mm/page_table_check.c b/mm/page_table_check.c
index 6ffc536359cd..ed3e1a76f266 100644
--- a/mm/page_table_check.c
+++ b/mm/page_table_check.c
@@ -278,7 +278,7 @@ void __page_table_check_pte_clear_range(struct mm_struct *mm,
 	if (&init_mm == mm)
 		return;
 
-	if (!pmd_bad(pmd) && !pmd_leaf(pmd)) {
+	if (!pmd_none(pmd) && !pmd_bad(pmd) && !pmd_leaf(pmd)) {
 		pte_t *ptep = pte_offset_map(&pmd, addr);
 		unsigned long i;
 
-- 
2.25.1


_______________________________________________
linux-riscv mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-riscv