[PATCH] locked page handling in shrink_cache()

[alad@hss.hns.com]

[-- Attachment #1: Type: text/plain, Size: 451 bytes --]



Whenever the shrink_cache wakes up after a laundered page in unlocked, it should
move that page to the end of inactive list so that
the page can be freed immediately and shrink_cache dosen't have to wait for
complete list scan before freeing this page.

The patch is created against standard redhat 7.1 distribution 2.4.16 kernel.
Let me know if I need to create it for a different kernel release.

Regards
Amol
(See attached file: vmscan.c-2.4.16)

[-- Attachment #2: vmscan.c-2.4.16 --]
[-- Type: application/octet-stream, Size: 288 bytes --]

--- vmscan_orig.c	Mon Jan  7 11:47:43 2002
+++ vmscan.c	Mon Jan  7 11:49:07 2002
@@ -387,6 +387,8 @@
 				wait_on_page(page);
 				page_cache_release(page);
 				spin_lock(&pagemap_lru_lock);
+				list_del(page->lru);
+				list_add(page->lru,inactive_list.prev);
 			}
 			continue;
 		}