Re: vm related lock up in 2.6.15

[Dave Jones <davej@redhat.com>]

On Mon, Jan 09, 2006 at 08:32:20AM -0800, Linus Torvalds wrote:
 
 > >  Turns out this was your BUG_ON(page_count(p) <= page_mapcount(p));
 > > addition to putpage_test_zero, which I added a few days ago and promptly
 > > forgot about.
 > > 
 > > Triggering this isn't too difficult it seems :-/
 > 
 > Well, as I warned in the message that had the patch, the test _is_ racy. 
 > The reads of the page counts have no serialization, so if another process 
 > is changing them, I don't guarantee that the test is correct.
 > 
 > IOW, it was meant as a special case debug test for one particular problem 
 > where I hoped it would give more information, rather than a real patch.

ok, that explains. I'll go back to chasing the real cause of the corruption
I was seeing :-/

		Dave