From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1754135AbZFUGYB (ORCPT ); Sun, 21 Jun 2009 02:24:01 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1752561AbZFUGXw (ORCPT ); Sun, 21 Jun 2009 02:23:52 -0400 Received: from atrey.karlin.mff.cuni.cz ([195.113.26.193]:46012 "EHLO atrey.karlin.mff.cuni.cz" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752181AbZFUGXw (ORCPT ); Sun, 21 Jun 2009 02:23:52 -0400 Date: Sun, 21 Jun 2009 08:18:47 +0200 From: Pavel Machek To: Benjamin Herrenschmidt Cc: Pekka J Enberg , linux-mm@kvack.org, linux-kernel@vger.kernel.org, mingo@elte.hu, npiggin@suse.de, akpm@linux-foundation.org, cl@linux-foundation.org, torvalds@linux-foundation.org, "Rafael J. Wysocki" Subject: Re: [PATCH v2] slab,slub: ignore __GFP_WAIT if we're booting or suspending Message-ID: <20090621061847.GB1474@ucw.cz> References: <20090619145913.GA1389@ucw.cz> <1245450449.16880.10.camel@pasglop> <20090619232336.GA2442@elf.ucw.cz> <1245455409.16880.15.camel@pasglop> <20090620002817.GA2524@elf.ucw.cz> <1245463809.16880.18.camel@pasglop> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <1245463809.16880.18.camel@pasglop> User-Agent: Mutt/1.5.18 (2008-05-17) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Hi! > > Academic for boot, probably real for suspend/resume. There the atomic > > reserves could matter because the memory can be pretty full when you > > start suspend. > > Right, that might be something to look into, though we haven't yet > applied the technique for suspend & resume. My main issue with it at the > moment is how do I synchronize with allocations that are already > sleeping when changing the gfp flag mask without bloating the normal Well, but the problem already exists, no? If someone is already sleeping due to __GFP_WAIT, he'll probably sleep till the resume. ...well, if he's sleeping in the disk driver, I suspect driver will finish outstanding requests as part of .suspend(). > I also suspect that we might want to try to make -some- amount of free > space before starting suspend, though of course not nearly as > aggressively as with std. We free 4MB in 2.6.30, but Rafael is removing that for 2.6.31 :-(. Pavel -- (english) http://www.livejournal.com/~pavelmachek (cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html