mm/thp: fix deadlock in collapse_file()

Commit Message

Song Liu Oct. 28, 2019, 10:14 p.m. UTC

As syzbot reported, we cannot call filemap_flush() with the page locked.
Remove the filemap_flush() as it is not required. khugepaged would just
wait until the page is flushed naturally.

Reported-by: [e-mail]
Fixes: 9d840e58caa0 ("mmthp-recheck-each-page-before-collapsing-file-thp-v4")
Cc: Johannes Weiner <[e-mail]>
Cc: Kirill A. Shutemov <[e-mail]>
Cc: Hugh Dickins <[e-mail]>
Cc: William Kucharski <[e-mail]>
Cc: Andrew Morton <[e-mail]>
Signed-off-by: Song Liu <[e-mail]>
---
 mm/khugepaged.c | 10 +---------
 1 file changed, 1 insertion(+), 9 deletions(-)

Patch

diff --git a/mm/khugepaged.c b/mm/khugepaged.c
index cd480dce92c6..3ec5333ae94d 100644
--- a/mm/khugepaged.c
+++ b/mm/khugepaged.c
@@ -1640,16 +1640,8 @@  static void collapse_file(struct mm_struct *mm,
 			/*
 			 * khugepaged only works on read-only fd, so this
 			 * page is dirty because it hasn't been flushed
-			 * since first write. There won't be new dirty
-			 * pages.
-			 *
-			 * Trigger async flush here and hope the writeback
-			 * is done when khugepaged revisits this page.
-			 *
-			 * This is a one-off situation. We are not forcing
-			 * writeback in loop.
+			 * since first write.
 			 */
-			filemap_flush(mapping);
 			result = SCAN_FAIL;
 			goto out_unlock;
 		}