Re: [PATCH 5.15] mm/filemap: fix UAF in find_lock_entries

[Matthew Wilcox]

On Wed, Jul 06, 2022 at 11:24:34AM +0800, Liu Shixin wrote:
> Release refcount after xas_set to fix UAF which may cause panic like this:

I think we can do better.  How about this?

diff --git a/mm/filemap.c b/mm/filemap.c
index 00e391e75880..11ae38cc4fd3 100644
--- a/mm/filemap.c
+++ b/mm/filemap.c
@@ -2090,7 +2090,9 @@ unsigned find_lock_entries(struct address_space *mapping, pgoff_t start,
 
 	rcu_read_lock();
 	while ((page = find_get_entry(&xas, end, XA_PRESENT))) {
+		unsigned long next_idx = xas.xa_index + 1;
 		if (!xa_is_value(page)) {
+			next_idx = page->index + thp_nr_pages(page);
 			if (page->index < start)
 				goto put;
 			if (page->index + thp_nr_pages(page) - 1 > end)
@@ -2111,14 +2113,11 @@ unsigned find_lock_entries(struct address_space *mapping, pgoff_t start,
 put:
 		put_page(page);
 next:
-		if (!xa_is_value(page) && PageTransHuge(page)) {
-			unsigned int nr_pages = thp_nr_pages(page);
-
-			/* Final THP may cross MAX_LFS_FILESIZE on 32-bit */
-			xas_set(&xas, page->index + nr_pages);
-			if (xas.xa_index < nr_pages)
-				break;
-		}
+		/* Final THP may cross MAX_LFS_FILESIZE on 32-bit */
+		if (next_idx < xas.xa_index)
+			break;
+		if (next_idx != xas.xa_index + 1)
+			xas_set(&xas, next_idx);
 	}
 	rcu_read_unlock();