nilfs2: prevent general protection fault in nilfs_clear_dirty_page()
authorRyusuke Konishi <konishi.ryusuke@gmail.com>
Mon, 12 Jun 2023 02:14:56 +0000 (11:14 +0900)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Wed, 28 Jun 2023 09:12:27 +0000 (11:12 +0200)
commit 782e53d0c14420858dbf0f8f797973c150d3b6d7 upstream.

In a syzbot stress test that deliberately causes file system errors on
nilfs2 with a corrupted disk image, it has been reported that
nilfs_clear_dirty_page() called from nilfs_clear_dirty_pages() can cause a
general protection fault.

In nilfs_clear_dirty_pages(), when looking up dirty pages from the page
cache and calling nilfs_clear_dirty_page() for each dirty page/folio
retrieved, the back reference from the argument page to "mapping" may have
been changed to NULL (and possibly others).  It is necessary to check this
after locking the page/folio.

So, fix this issue by not calling nilfs_clear_dirty_page() on a page/folio
after locking it in nilfs_clear_dirty_pages() if the back reference
"mapping" from the page/folio is different from the "mapping" that held
the page/folio just before.

Link: https://lkml.kernel.org/r/20230612021456.3682-1-konishi.ryusuke@gmail.com
Signed-off-by: Ryusuke Konishi <konishi.ryusuke@gmail.com>
Reported-by: syzbot+53369d11851d8f26735c@syzkaller.appspotmail.com
Closes: https://lkml.kernel.org/r/000000000000da4f6b05eb9bf593@google.com
Tested-by: Ryusuke Konishi <konishi.ryusuke@gmail.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
fs/nilfs2/page.c

index 39b7eea..7d31833 100644 (file)
@@ -369,7 +369,15 @@ void nilfs_clear_dirty_pages(struct address_space *mapping, bool silent)
                        struct page *page = pvec.pages[i];
 
                        lock_page(page);
-                       nilfs_clear_dirty_page(page, silent);
+
+                       /*
+                        * This page may have been removed from the address
+                        * space by truncation or invalidation when the lock
+                        * was acquired.  Skip processing in that case.
+                        */
+                       if (likely(page->mapping == mapping))
+                               nilfs_clear_dirty_page(page, silent);
+
                        unlock_page(page);
                }
                pagevec_release(&pvec);