Re: mm/rmap.c negative page map count BUG.

[Dave Jones <davej@redhat.com>]

On Wed, Jan 04, 2006 at 03:53:26PM -0800, Andrew Morton wrote:
 > Dave Jones <davej@redhat.com> wrote:
 > >
 > >  > Can you print ->flags, ->count, ->mapping, etc instead of going BUG?
 > > 
 > > I can add some instrumentation like this though, and see what turns up.
 > 
 > Can we get that instrumentation into the upstream kernel please?  We do
 > seem to be hitting rmap assertions too often for it to be dud
 > hardware/bodgy drivers/etc.

This is what I came up with..
anything missing ?

		Dave

--- linux-2.6.14/mm/rmap.c~	2006-01-03 08:53:32.000000000 -0500
+++ linux-2.6.14/mm/rmap.c	2006-01-03 08:58:19.000000000 -0500
@@ -484,6 +484,13 @@ void page_remove_rmap(struct page *page)
 	BUG_ON(PageReserved(page));
 
 	if (atomic_add_negative(-1, &page->_mapcount)) {
+		if (page_mapcount(page) < 0) {
+			printk (KERN_EMERG "Eeek! page_mapcount(page) went negative! (%d)\n", page->_mapcount);
+			printk (KERN_EMERG "  page->flags = %x\n", page->flags);
+			printk (KERN_EMERG "  page->count = %x\n", page->_count);
+			printk (KERN_EMERG "  page->mapping = %p\n", page->mapping);
+		}
+		
 		BUG_ON(page_mapcount(page) < 0);
 		/*
 		 * It would be tidy to reset the PageAnon mapping here,