From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1750789AbWBCNxk (ORCPT ); Fri, 3 Feb 2006 08:53:40 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1750791AbWBCNxk (ORCPT ); Fri, 3 Feb 2006 08:53:40 -0500 Received: from silver.veritas.com ([143.127.12.111]:41794 "EHLO silver.veritas.com") by vger.kernel.org with ESMTP id S1750789AbWBCNxj (ORCPT ); Fri, 3 Feb 2006 08:53:39 -0500 Date: Fri, 3 Feb 2006 13:54:20 +0000 (GMT) From: Hugh Dickins X-X-Sender: hugh@goblin.wat.veritas.com To: Badri Pillai cc: Martin Drab , linux-kernel@vger.kernel.org Subject: Re: PROBLEM [ATI fglrx 2.6.15 put_page BUG] In-Reply-To: <43E34662.1000704@ii-consult.com> Message-ID: References: <43E34662.1000704@ii-consult.com> MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII X-OriginalArrivalTime: 03 Feb 2006 13:53:39.0173 (UTC) FILETIME=[3C4B1950:01C628C9] Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org On Fri, 3 Feb 2006, Badri Pillai wrote: > > Problem summary: after upgrading to from 2.6.15.1 -> 2.6.15.2 > the system dumps following trace, when trying to reboot through KDE 3.5: I doubt it's anything new in 2.6.15.2. > - ------------[ cut here ]------------ > kernel BUG at mm/swap.c:49! > invalid operand: 0000 [#2] That's #2 so there should have been an earlier one, but never mind... > SMP > Modules linked in: nls_utf8 ntfs snd_rtctimer ipv6 fglrx .... ATI's fglrx is Tainting your kernel. > EIP: 0060:[] Tainted: P B VLI > EFLAGS: 00013256 (2.6.15.2_DL6) I don't know 2.6.15.2_DL6, but never mind... > EIP is at put_page+0x4a/0x66 > Process X (pid: 1682, threadinfo=db55e000 task=c17fe570) And further down... > [] put_page+0x4a/0x66 > [] avc_has_perm_noaudit+0x37/0xe3 > [] drm_free+0xb7/0x150 [fglrx] > [] drm_free+0xb7/0x150 [fglrx] > [] error_code+0x4f/0x54 > [] FGLDRM__vm_info+0xa0/0x16b [fglrx] > [] FGLDRM__vm_info+0xa0/0x16b [fglrx] > [] put_page+0x4a/0x66 > [] zap_pte_range+0x172/0x236 > [] unmap_page_range+0xa9/0xf8 > [] unmap_vmas+0x101/0x1e2 > [] unmap_region+0x92/0xf8 > [] do_munmap+0xd9/0xef > [] sys_munmap+0x50/0x68 > [] sysenter_past_esp+0x54/0x75 Yes, fglrx definitely looks implicated. ATI were carefully coding around an anomaly in our page count handling, then 2.6.15 removed that anomaly, so their wrapper code needs an additional #ifdef on Linux kernel version. I'll forward you offline what Martin Drab posted a few weeks back. Hugh