From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S261389AbUDCByd (ORCPT ); Fri, 2 Apr 2004 20:54:33 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S261425AbUDCByd (ORCPT ); Fri, 2 Apr 2004 20:54:33 -0500 Received: from ppp-217-133-42-200.cust-adsl.tiscali.it ([217.133.42.200]:6530 "EHLO dualathlon.random") by vger.kernel.org with ESMTP id S261389AbUDCByb (ORCPT ); Fri, 2 Apr 2004 20:54:31 -0500 Date: Sat, 3 Apr 2004 03:54:27 +0200 From: Andrea Arcangeli To: linux-kernel@vger.kernel.org Subject: 2.6.5-rc3-aa3 Message-ID: <20040403015427.GA2307@dualathlon.random> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline User-Agent: Mutt/1.4.1i X-GPG-Key: 1024D/68B9CB43 13D9 8355 295F 4823 7C49 C012 DFA1 686E 68B9 CB43 X-PGP-Key: 1024R/CB4660B9 CC A0 71 81 F4 A0 63 AC C0 4B 81 1D 8C 15 C8 E5 Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org this avoids (harmless) warnings while freeing non-compound pages, adds some more robustness check to the freelist by making sure no page has PG_compound set, and most important it adds merging for filebacked mappings and anonymous mappings in mprotect. I'm running it on my main desktop and there are no problems so far. I believe the VM side is feature complete, in order: ----------- [..] # -aa VM (remove rmap) objrmap-core anon-vma prio-tree gfp-no-compound mprotect-vma-merging ----------- Two pending thing to clarify are swapsuspend/swapresume (works partially for me because aic7xxx isn't swap-resume capable), and Christoph's xfs oops on ppc that looks a miscompilation of some sort. URL: http://www.us.kernel.org/pub/linux/kernel/people/andrea/kernels/v2.6/2.6.5-rc3-aa3.gz http://www.us.kernel.org/pub/linux/kernel/people/andrea/kernels/v2.6/2.6.5-rc3-aa3/ Changelog diff between 2.6.5-rc3-aa2 and 2.6.5-rc3-aa3: Files 2.6.5-rc3-aa2/extraversion and 2.6.5-rc3-aa3/extraversion differ Rediffed. Files 2.6.5-rc3-aa2/gfp-no-compound and 2.6.5-rc3-aa3/gfp-no-compound differ Avoid flood of warnings while freeing a non-compound multipage. Added bugchecks to be sure all pages in the freelist have PG_compound unset (for robusteness, the check for PG_compound is zerocost anyways since we check it in a bitmask). Christoph's troubles with ppc seems to be a gcc miscompilation if it's really the second bad_page in destroy_compound_page triggering. The swap suspend/swap resume works for me w/o apparent VM issues, aic7xxx simply hangs at resume, but Pavel said it's expected. I'll later try to load it on my laptop and see if I've more luck with it. Only in 2.6.5-rc3-aa3: mprotect-vma-merging Resurrected mprotect anon-vma and file-backed merging. The file-backed merging is IMHO the most interesting one, and it's for the first time available in linux. Files 2.6.5-rc3-aa2/prio-tree.gz and 2.6.5-rc3-aa3/prio-tree.gz differ Added s390 compilation fix.