From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1754619Ab0LGCCr (ORCPT ); Mon, 6 Dec 2010 21:02:47 -0500 Received: from cerberus.infoblox.com ([128.242.99.209]:56176 "EHLO cerberus.infoblox.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754179Ab0LGCCq (ORCPT ); Mon, 6 Dec 2010 21:02:46 -0500 X-Greylist: delayed 2026 seconds by postgrey-1.27 at vger.kernel.org; Mon, 06 Dec 2010 21:02:46 EST Message-ID: <4CFD8DD9.1000101@infoblox.com> Date: Mon, 6 Dec 2010 17:28:57 -0800 From: Kin Cho User-Agent: Mozilla/5.0 (X11; U; Linux x86_64; en-US; rv:1.9.1.12) Gecko/20100907 Fedora/3.0.7-1.fc12 Thunderbird/3.0.7 MIME-Version: 1.0 To: Subject: [x86_64] "bad pmd" when using huge pages Content-Type: text/plain; charset="ISO-8859-1"; format=flowed Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Hi, With a stock 2.6.32-8 kernel (with minor patches) configured for x86_64 and using huge pages, I'm seeing dmesg output like this: mm/memory.c:144: bad pmd ffff880117bb6128(00000002cca000e7). At this point one huge page is apparently "leaked" -- h->free_huge_pages in gather_surplus_pages is 1 less, until rebooted. When this happens, the system is moderately loaded (~70% idle). Also, if processes (bdb) holding shared (huge) pages are not restarted, or when configured for x86, the problem doesn't appear. Looking at the pmd flags, the problem appears to be related to the _PAGE_PSE bit incorrectly getting set. Any suggestions/comments is appreciated. Thanks, -kin