Re: VM-related Oops: 2.4.15pre1

[Christian Ehrhardt]

On Fri, Nov 16, 2001 at 02:23:44PM -0800, Simon Kirby wrote:
> This box has Oopsed twice but is still running.  Both Oopses followed a
> BUG() report (same in both cases):
> 
> kernel BUG at page_alloc.c:76!
> 
> Which maps to:
> 
>         if (page->mapping)
> 		BUG();
> 
> ...in __free_pages_ok() in mm/page_alloc.c.

I think this one liner (diffed against 2.4.14) could fix this Oops:

--- mm/vmscan.c.vanilla	Sat Nov 17 23:37:01 2001
+++ mm/vmscan.c	Sat Nov 17 23:39:04 2001
@@ -414,10 +414,9 @@
 		 * the page as well.
 		 */
 		if (page->buffers) {
-			spin_unlock(&pagemap_lru_lock);
-
 			/* avoid to free a locked page */
 			page_cache_get(page);
+			spin_unlock(&pagemap_lru_lock);
 
 			if (try_to_free_buffers(page, gfp_mask)) {
 				if (!page->mapping) {

    regards   Christian

-- 
THAT'S ALL FOLKS!