From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1755850AbYBAIY3 (ORCPT ); Fri, 1 Feb 2008 03:24:29 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1752339AbYBAIYV (ORCPT ); Fri, 1 Feb 2008 03:24:21 -0500 Received: from rv-out-0910.google.com ([209.85.198.185]:25921 "EHLO rv-out-0910.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1750867AbYBAIYU (ORCPT ); Fri, 1 Feb 2008 03:24:20 -0500 DomainKey-Signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma; h=message-id:date:from:to:subject:cc:in-reply-to:mime-version:content-type:content-transfer-encoding:content-disposition:references; b=e9NVSe4FA9dP56JswL3tWVIFbw7tGQX+0Su1HdRnpdCt32GUISzv/rj3R6FPVewATFCd/D24MiPxAhQ7s3535pvyiwx6XSCEC6oxg7lNNobkUswNM54mxtziS1JDgij9kRUdyoquiyMOjV0wbs0A3r23sRWxEz8y/LUOo9uW0Lg= Message-ID: <3fd7d7a70802010024q22b4d179mf56e6d4b60e4f574@mail.gmail.com> Date: Fri, 1 Feb 2008 17:24:17 +0900 From: "Kenichi Okuyama" To: "Andrew Morton" Subject: Re: [patch] NULL pointer check for vma->vm_mm Cc: linux-mm@kvack.org, linux-kernel@vger.kernel.org In-Reply-To: <20080131235544.346b938a.akpm@linux-foundation.org> MIME-Version: 1.0 Content-Type: text/plain; charset=ISO-2022-JP Content-Transfer-Encoding: 7bit Content-Disposition: inline References: <3fd7d7a70801312339p2a142096p83ed286c81379728@mail.gmail.com> <20080131235544.346b938a.akpm@linux-foundation.org> Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Dear Andrew, all, First of all, thank you for looking at the patch. I do agree that if mm is NULL, system will call Oops anyway. However, since it's oops, it does not stop the system, nor call kdump. By calling BUG_ON(), it'll gives us chance of calling kdump at the first chance. Since this is very rare to happen, I thought we should capture the incident whenever possible. On other hand, because BUG_ON macro is very light, I thought this will not harm any performance... Forgive me in advance if I was wrong. I still think checking mm with BUG_ON here is better than counting on Oops. best regards, 2008/2/1, Andrew Morton : > On Fri, 1 Feb 2008 16:39:07 +0900 "Kenichi Okuyama" wrote: > > > Dear all, > > > > I was looking at the ./mm/rmap.c .. I found that, in function > > "page_referenced_one()", > > struct mm_struct *mm = vma->vm_mm; > > was being refererred without NULL check. > > > > Though I do agree that this works for most of the cases, I thought it > > is better to add > > BUG_ON() for case of mm being NULL. > > > > attached is the patch for this > > If we dereference NULL then the kernel will display basically the same > information as would a BUG, and it takes the same action. So adding a > BUG_ON here really doesn't gain us anything. > > Also, I think vma->vm_mm == 0 is not a valid state, so this just shouldn't > happen - the code is OK to assume that a particular invariant is being > honoured. > > -- 奥山　健一(Kenichi Okuyama) [煤背会: No. 0x00000001] URL: http://www.dd.iij4u.or.jp/~okuyamak/ http://developer.osdl.jp/projects/doubt/