mm/huge_memory: fix folio_set_dirty() vs. folio_mark_dirty()
authorDavid Hildenbrand <david@redhat.com>
Mon, 22 Jan 2024 17:54:07 +0000 (18:54 +0100)
committerAndrew Morton <akpm@linux-foundation.org>
Fri, 26 Jan 2024 07:52:21 +0000 (23:52 -0800)
The correct folio replacement for "set_page_dirty()" is
"folio_mark_dirty()", not "folio_set_dirty()".  Using the latter won't
properly inform the FS using the dirty_folio() callback.

This has been found by code inspection, but likely this can result in some
real trouble.

Link: https://lkml.kernel.org/r/20240122175407.307992-1-david@redhat.com
Fixes: a8e61d584eda0 ("mm/huge_memory: page_remove_rmap() -> folio_remove_rmap_pmd()")
Signed-off-by: David Hildenbrand <david@redhat.com>
Cc: Ryan Roberts <ryan.roberts@arm.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
mm/huge_memory.c

index 94ef5c0..5f31e51 100644 (file)
@@ -2437,7 +2437,7 @@ static void __split_huge_pmd_locked(struct vm_area_struct *vma, pmd_t *pmd,
                        page = pmd_page(old_pmd);
                        folio = page_folio(page);
                        if (!folio_test_dirty(folio) && pmd_dirty(old_pmd))
-                               folio_set_dirty(folio);
+                               folio_mark_dirty(folio);
                        if (!folio_test_referenced(folio) && pmd_young(old_pmd))
                                folio_set_referenced(folio);
                        folio_remove_rmap_pmd(folio, page, vma);
@@ -3563,7 +3563,7 @@ int set_pmd_migration_entry(struct page_vma_mapped_walk *pvmw,
        }
 
        if (pmd_dirty(pmdval))
-               folio_set_dirty(folio);
+               folio_mark_dirty(folio);
        if (pmd_write(pmdval))
                entry = make_writable_migration_entry(page_to_pfn(page));
        else if (anon_exclusive)