This patch fixes oops for bz#202962, bz#198698, bz#141703.

On s390, pages mapped by kernel get dirty when kernel stores into them.
SetPageUptodate() clears the dirty bit, but only if the page was not
up to date before. Thus, when it is called twice on the same page,
the dirty bit is not cleared the second time around. This produces
a erroneous dirty page and an oops.

There's a more detailed explanation here:
 http://marc.theaimsgroup.com/?t=115725548100001&r=1&w=2
As always, Bugzilla has all the materials too.

I tested this patch to fix the problem. I did not test squashfs to
continue to work on all other platforms though... It seems like working.

Squashfs is not upstream in Linus or Andrew's trees. We use it because
of need (I asked Jeremy). The upstream maintainer (Phillip Lougher) liked
the patch, but there was no "upstream" testing yet.

Please review and ack. Especially I'm looking at those who knows what
SetPageUptodate() and grab_cache_page_nowait() are.

-- Pete

diff -urp -X dontdiff linux-2.6.17-1.2519.4.5.el5/fs/squashfs/inode.c linux-2.6.17-1.2519.4.5.el5.z1/fs/squashfs/inode.c
--- linux-2.6.17-1.2519.4.5.el5/fs/squashfs/inode.c	2006-08-25 01:44:10.000000000 -0400
+++ linux-2.6.17-1.2519.4.5.el5.z1/fs/squashfs/inode.c	2006-09-01 22:33:24.000000000 -0400
@@ -1554,8 +1554,15 @@ static int squashfs_readpage(struct file
 			flush_dcache_page(page);
 			SetPageUptodate(page);
 			unlock_page(page);
-		} else if ((push_page =
-				grab_cache_page_nowait(page->mapping, i))) {
+		} else {
+			push_page = grab_cache_page_nowait(page->mapping, i);
+			if (!push_page)
+				continue;
+			if (PageUptodate(push_page)) {
+				unlock_page(push_page);
+				page_cache_release(push_page);
+				continue;
+			}
  			pageaddr = kmap_atomic(push_page, KM_USER0);
 
 			memcpy(pageaddr, data_ptr + byte_offset,