[PATCH] mm: free_pages_and_swap_cache opt

Minor optimization (though it doesn't help in the PREEMPT case, severely
constrained by small ZAP_BLOCK_SIZE).  free_pages_and_swap_cache works in
chunks of 16, calling release_pages which works in chunks of PAGEVEC_SIZE.
But PAGEVEC_SIZE was dropped from 16 to 14 in 2.6.10, so we're now doing more
spin_lock_irq'ing than necessary: use PAGEVEC_SIZE throughout.

Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Nick Piggin <nickpiggin@yahoo.com.au>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

mm/swap_state.c

@@ -14,6 +14,7 @@
 #include <linux/pagemap.h>
 #include <linux/buffer_head.h>
 #include <linux/backing-dev.h>
+#include <linux/pagevec.h>
 
 #include <asm/pgtable.h>
 
@@ -272,12 +273,11 @@ void free_page_and_swap_cache(struct page *page)
  */
 void free_pages_and_swap_cache(struct page **pages, int nr)
 {
-	int chunk = 16;
 	struct page **pagep = pages;
 
 	lru_add_drain();
 	while (nr) {
-		int todo = min(chunk, nr);
+		int todo = min(nr, PAGEVEC_SIZE);
 		int i;
 
 		for (i = 0; i < todo; i++)