[PATCH v4 14/18] f2fs: wait for folio writeback before ffs_detach_free

[Nanzhe Zhao <nzzhao.sigma@gmail.com>]

From: Nanzhe Zhao <zhaonanzhe@xiaomi.com>

truncate_inode_pages_range() can invalidate a large folio that is still
under writeback, which can result in the following use-after-free:

 CPU0 (writeback)             CPU1 (truncate)
 ----                         ----
f2fs_write_cache_folios()
folio_start_writeback(folio)
/* submit bios */

retry_out:
folio_unlock(folio)
/* all bios submitted; folio lock released */

                              truncate_inode_pages_range()
                              find_lock_entries() /* locks the folio */
                              truncate_cleanup_folio(folio)
                              folio_invalidate(folio, 0, folio_size(folio))
                              -> f2fs_invalidate_folio(folio, 0,
                                                        folio_size(folio))

f2fs_write_end_bio(bio)
ffs = folio->private          ffs_detach_free(folio)
                              folio_detach_private(folio)
                              kfree(ffs)
atomic_sub_and_test(nr_pages,
    &ffs->write_pages_pending) <-- use-after-free

Wait for folio writeback in invalidate_folio, then we can free the
folio state without any other remaining user.

Reported-by: kernel test robot <yi1.lai@intel.com>
Closes: https://lore.kernel.org/oe-lkp/202610081027.b13dcf10-lkp@intel.com
Signed-off-by: Nanzhe Zhao <zhaonanzhe@xiaomi.com>
---
 fs/f2fs/data.c | 8 ++++++++
 1 file changed, 8 insertions(+)

diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c
index 61fe4a0c5..433d8f2b0 100644
--- a/fs/f2fs/data.c
+++ b/fs/f2fs/data.c
@@ -5427,6 +5427,14 @@ void f2fs_invalidate_folio(struct folio *folio, size_t offset, size_t length)
 	}
 
 	if (f2fs_folio_has_ffs(folio)) {
+		/*
+		 * The folio may still be under writeback with bios in
+		 * completion that need to access f2fs_folio_state.  Wait
+		 * for the writeback to complete before freeing the state,
+		 * otherwise it will result in use-after-free.
+		 */
+		folio_wait_writeback(folio);
+
 		f2fs_ffs_clear_subrange_dirty_and_test(folio, 0,
 							folio_size(folio));
 	} else if (folio_test_dirty(folio)) {
-- 
2.43.0