From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1758355Ab2DJCZ6 (ORCPT ); Mon, 9 Apr 2012 22:25:58 -0400 Received: from beauty.rexursive.com ([150.101.121.179]:48325 "EHLO beauty.rexursive.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753016Ab2DJCZ5 (ORCPT ); Mon, 9 Apr 2012 22:25:57 -0400 Message-ID: <1334024754.2103.9.camel@shrek.rexursive.com> Subject: Re: [PATCH v8]: Hibernation: lower/better control the amount of pages used for buffering From: Bojan Smojver To: "Rafael J. Wysocki" Cc: linux-kernel@vger.kernel.org, Linux PM list , Per Olofsson Date: Tue, 10 Apr 2012 12:25:54 +1000 In-Reply-To: <2cbc3801-3ab0-4ef6-84cd-d69a23ff4dcc@email.android.com> References: <1333418527.2035.11.camel@shrek.rexursive.com> <201204030631.24974.rjw@sisk.pl> <1333432393.2035.13.camel@shrek.rexursive.com> <1333620908.8191.1.camel@shrek.rexursive.com> <2cbc3801-3ab0-4ef6-84cd-d69a23ff4dcc@email.android.com> Content-Type: text/plain; charset="UTF-8" X-Mailer: Evolution 3.2.3 (3.2.3-2.fc16) Content-Transfer-Encoding: 7bit Mime-Version: 1.0 Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Sun, 2012-04-08 at 10:14 +1000, Bojan Smojver wrote: > Please ignore this request. As it turns out, the hangs are still > there, it just takes a bit longer to trigger. Thanks to Per Olofsson, who actually determined what was going on, we now know what the problem is. My "free pages" maths used nr_free_pages() macro, which deals with all free pages. However, we only allocate buffers from non-high zones. So, on systems where there is a significant number of high pages free, we get a very large number (even if only a quarter is used), which then gets the kernel into a bind and eventually hangs it on page allocation. Will send a patch shortly, which will also have some other fixes. -- Bojan