Re: PATCH: Further aacraid work

[William Lee Irwin III <wli@holomorphy.com>]

On Thu, Jun 17, 2004 at 01:54:38PM -0400, Salyzyn, Mark wrote:
>> And I might add, undoing the entropy to result in the descending page
>> list (but that is the forth time I've said this).
>> I ran heavy sequential load overnight and continued to have this
>> characteristic when taking snapshots of command SG lists. The average SG
>> element size statistically was 4168 bytes.

On Thu, Jun 17, 2004 at 04:38:42PM -0400, Alan Cox wrote:
> What do the stats look like with the patch Andrew Morton (I think) posted
> to reverse the page order from the allocator ?

Say, could you guys try this? jejb seemed to get decent results with it.


===== mm/page_alloc.c 1.211 vs edited =====
--- 1.211/mm/page_alloc.c	Sat Jun 12 20:52:26 2004
+++ edited/mm/page_alloc.c	Thu Jun 17 13:46:36 2004
@@ -297,14 +297,12 @@
 	unsigned long size = 1 << high;
 
 	while (high > low) {
-		BUG_ON(bad_range(zone, page));
 		area--;
 		high--;
 		size >>= 1;
-		list_add(&page->lru, &area->free_list);
-		MARK_USED(index, high, area);
-		index += size;
-		page += size;
+		BUG_ON(bad_range(zone, &page[size]));
+		list_add(&page[size].lru, &area->free_list);
+		MARK_USED(index + size, high, area);
 	}
 	return page;
 }