[PATCH] valid_swaphandles() fix

akpm draws my attention to the fact that sysctl(VM_PAGE_CLUSTER) might
conceivably change page_cluster to 0 while valid_swaphandles() is in the
middle of using it, leading to an embarrassingly long loop: take a local
snapshot of page_cluster and work with that.
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

mm/swapfile.c

@@ -1723,13 +1723,14 @@ get_swap_info_struct(unsigned type)
  */
 int valid_swaphandles(swp_entry_t entry, unsigned long *offset)
 {
-	int ret = 0, i = 1 << page_cluster;
+	int our_page_cluster = page_cluster;
+	int ret = 0, i = 1 << our_page_cluster;
 	unsigned long toff;
 	struct swap_info_struct *swapdev = swp_type(entry) + swap_info;
 
-	if (!page_cluster)	/* no readahead */
+	if (!our_page_cluster)	/* no readahead */
 		return 0;
-	toff = (swp_offset(entry) >> page_cluster) << page_cluster;
+	toff = (swp_offset(entry) >> our_page_cluster) << our_page_cluster;
 	if (!toff)		/* first page is swap header */
 		toff++, i--;
 	*offset = toff;