From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mail172.messagelabs.com (mail172.messagelabs.com [216.82.254.3]) by kanga.kvack.org (Postfix) with ESMTP id 7062B6B0047 for ; Tue, 10 Mar 2009 06:49:50 -0400 (EDT) Date: Tue, 10 Mar 2009 11:49:47 +0100 From: Nick Piggin Subject: Re: possible bug in find_get_pages Message-ID: <20090310104947.GB4594@wotan.suse.de> References: <20090306192625.GA3267@linux.intel.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: Sender: owner-linux-mm@kvack.org To: Christoph Lameter Cc: mark gross , linux-mm@kvack.org List-ID: On Fri, Mar 06, 2009 at 02:28:50PM -0500, Christoph Lameter wrote: > On Fri, 6 Mar 2009, mark gross wrote: > > > It seems that page->_count == 0 at some point on some overnight runs > > with locks the system into a tight loop from the repeat: and a goto > > repeat in find_get_pages. > > A page with ref count zero should not be in any mapping. If the page is in > a mapping then the page is used. Therefore the refcount should be > 0. > > If there is a page with zero refcount and its in a mapping then something > erroneously decreased the refcount. Just for posterity, this isn't _quite_ true any more with Hugh's variation to the speculative reference method. We now in some places set the page's refcount to 0 in order to hold off new speculative references from turning into real references (eg. right before final checks before page reclaim). But yes, such a page should not remain both in a mapping and with a 0 refcount for long periods. -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@kvack.org. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: email@kvack.org