Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1752122AbdGRUCC (ORCPT ); Tue, 18 Jul 2017 16:02:02 -0400 Received: from mail-oi0-f67.google.com ([209.85.218.67]:36732 "EHLO mail-oi0-f67.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751457AbdGRUCA (ORCPT ); Tue, 18 Jul 2017 16:02:00 -0400 MIME-Version: 1.0 In-Reply-To: References: <20170714092540.1217397-1-arnd@arndb.de> <20170714092540.1217397-8-arnd@arndb.de> From: Arnd Bergmann Date: Tue, 18 Jul 2017 22:01:59 +0200 X-Google-Sender-Auth: sedeJr0usdZWk3eNlfFamzZqr_E Message-ID: Subject: Re: [PATCH 07/14] proc/kcore: hide a harmless warning To: Ard Biesheuvel Cc: "linux-kernel@vger.kernel.org" , Jiri Olsa , Greg Kroah-Hartman , Linus Torvalds , Tejun Heo , Guenter Roeck , IDE-ML , Linux Media Mailing List , Andrew Morton , dri-devel , Kees Cook , Ingo Molnar , Laura Abbott , Pratyush Anand Content-Type: text/plain; charset="UTF-8" Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Transfer-Encoding: 8bit X-MIME-Autoconverted: from quoted-printable to 8bit by nfs id v6IK287A029832 Content-Length: 2592 Lines: 64 On Tue, Jul 18, 2017 at 9:55 PM, Ard Biesheuvel wrote: > On 18 July 2017 at 20:53, Arnd Bergmann wrote: >> On Fri, Jul 14, 2017 at 2:28 PM, Ard Biesheuvel >> wrote: >>> On 14 July 2017 at 10:25, Arnd Bergmann wrote: >>>> gcc warns when MODULES_VADDR/END is defined to the same value as >>>> VMALLOC_START/VMALLOC_END, e.g. on x86-32: >>>> >>>> fs/proc/kcore.c: In function ‘add_modules_range’: >>>> fs/proc/kcore.c:622:161: error: self-comparison always evaluates to false [-Werror=tautological-compare] >>>> if (/*MODULES_VADDR != VMALLOC_START && */MODULES_END != VMALLOC_END) { >>>> >>> >>> Does it occur for subtraction as well? Or only for comparison? >> >> This replacement patch would also address the warning: >> >> diff --git a/fs/proc/kcore.c b/fs/proc/kcore.c >> index 45629f4b5402..35824e986c2c 100644 >> --- a/fs/proc/kcore.c >> +++ b/fs/proc/kcore.c >> @@ -623,7 +623,7 @@ static void __init proc_kcore_text_init(void) >> struct kcore_list kcore_modules; >> static void __init add_modules_range(void) >> { >> - if (MODULES_VADDR != VMALLOC_START && MODULES_END != VMALLOC_END) { >> + if (MODULES_VADDR - VMALLOC_START && MODULES_END - VMALLOC_END) { >> kclist_add(&kcore_modules, (void *)MODULES_VADDR, >> MODULES_END - MODULES_VADDR, KCORE_VMALLOC); >> } >> >> I have also verified that four of the 14 patches are not needed when building >> without ccache, this is one of them: >> >> acpi: thermal: fix gcc-6/ccache warning >> proc/kcore: hide a harmless warning >> SFI: fix tautological-compare warning >> [media] fix warning on v4l2_subdev_call() result interpreted as bool >> >> Not sure what to do with those, we could either ignore them all and >> not care about ccache, or we try to address them all in some way. >> > > Any idea why ccache makes a difference here? It is not obvious (not to > me at least) When ccache is used, the compilation stage is apparently always done on the preprocessed source file. So instead of parsing (with the integrated preprocessor) if (MODULES_VADDR != VMALLOC_START ...) the compiler sees if (((unsigned long)high_memory + (8 * 1024 * 1024)) != ((unsigned long)high_memory + (8 * 1024 * 1024)) ...) and it correctly considers the first expression something that one would write in source code, while -Wtautological-compare is intended to warn about the second version being always true, which makes the 'if()' pointless. Arnd