page_launder() bug

From: BERECZ Szabolcs (szabi@inf.elte.hu)
Date: Sun May 06 2001 - 16:08:20 EST


Hi!

there is a bug in page_launder introduced with kernel 2.4.3-ac12.
if the swapfile is on a filesystem, then after swapping out some
pages, the system locks up. sometimes it writes an oops message.
I don't know exactly what's the problem, but with the attached
patch it works. (this is just a reverse patch to go back to
pre 2.4.3-ac12, and it's against 2.4.4-ac5)
please fix the bug, or apply this patch until fixed properly.

a question:
why don't you include lkcd or something like that in the
mainstream kernel? it would be much easier to save those annoying
oopses :)

Bye,
Szabi

diff -Nur linux/mm/vmscan.c linux.swapfix/mm/vmscan.c
--- linux/mm/vmscan.c Sun May 6 15:59:22 2001
+++ linux.swapfix/mm/vmscan.c Sun May 6 16:07:09 2001
@@ -448,15 +448,9 @@
         maxscan = nr_inactive_dirty_pages;
         while ((page_lru = inactive_dirty_list.prev) != &inactive_dirty_list &&
                                 maxscan-- > 0) {
- int dead_swap_page;
-
                 page = list_entry(page_lru, struct page, lru);
                 zone = page->zone;

- dead_swap_page =
- (PageSwapCache(page) &&
- page_count(page) == (1 + !!page->buffers));
-
                 /* Wrong page on list?! (list corruption, should not happen) */
                 if (!PageInactiveDirty(page)) {
                         printk("VM: page_launder, wrong page on list.\n");
@@ -467,10 +461,9 @@
                 }

                 /* Page is or was in use? Move it to the active list. */
- if (!dead_swap_page &&
- (PageTestandClearReferenced(page) || page->age > 0 ||
- (!page->buffers && page_count(page) > 1) ||
- page_ramdisk(page))) {
+ if (PageTestandClearReferenced(page) || page->age > 0 ||
+ (!page->buffers && page_count(page) > 1) ||
+ page_ramdisk(page)) {
                         del_page_from_inactive_dirty_list(page);
                         add_page_to_active_list(page);
                         continue;
@@ -512,11 +505,8 @@
                         if (!writepage)
                                 goto page_active;

- /* First time through? Move it to the back of the list,
- * but not if it is a dead swap page. We want to reap
- * those as fast as possible.
- */
- if (!launder_loop && !dead_swap_page) {
+ /* First time through? Move it to the back of the list */
+ if (!launder_loop) {
                                 list_del(page_lru);
                                 list_add(page_lru, &inactive_dirty_list);
                                 UnlockPage(page);

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/



This archive was generated by hypermail 2b29 : Mon May 07 2001 - 21:00:23 EST