From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <linux-kernel-owner@vger.kernel.org>
Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand
	id S1751887AbbHMCHG (ORCPT <rfc822;w@1wt.eu>);
	Wed, 12 Aug 2015 22:07:06 -0400
Received: from mx1.redhat.com ([209.132.183.28]:48797 "EHLO mx1.redhat.com"
	rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP
	id S1751160AbbHMCHE (ORCPT <rfc822;linux-kernel@vger.kernel.org>);
	Wed, 12 Aug 2015 22:07:04 -0400
Date: Wed, 12 Aug 2015 21:07:00 -0500
From: Josh Poimboeuf <jpoimboe@redhat.com>
To: Chris J Arges <chris.j.arges@canonical.com>
Cc: Thomas Gleixner <tglx@linutronix.de>, Ingo Molnar <mingo@redhat.com>,
        "H. Peter Anvin" <hpa@zytor.com>, Michal Marek <mmarek@suse.cz>,
        Peter Zijlstra <peterz@infradead.org>,
        Andy Lutomirski <luto@kernel.org>, Borislav Petkov <bp@alien8.de>,
        Linus Torvalds <torvalds@linux-foundation.org>,
        Andi Kleen <andi@firstfloor.org>, Pedro Alves <palves@redhat.com>,
        Namhyung Kim <namhyung@gmail.com>,
        Bernd Petrovitsch <bernd@petrovitsch.priv.at>, x86@kernel.org,
        live-patching@vger.kernel.org, linux-kernel@vger.kernel.org
Subject: Re: [PATCH v8 00/21] Compile-time stack validation
Message-ID: <20150813020700.GC25439@treble.redhat.com>
References: <55CA699B.4070809@canonical.com>
 <20150812044317.GB25439@treble.redhat.com>
 <55CBB9A1.7000800@canonical.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Disposition: inline
In-Reply-To: <55CBB9A1.7000800@canonical.com>
User-Agent: Mutt/1.5.23.1-rc1 (2014-03-12)
Sender: linux-kernel-owner@vger.kernel.org
List-ID: <linux-kernel.vger.kernel.org>
X-Mailing-List: linux-kernel@vger.kernel.org

On Wed, Aug 12, 2015 at 04:24:49PM -0500, Chris J Arges wrote:
> <snip>
> > 
> > Thanks for trying it out.  I couldn't figure out how to recreate this
> > exact error, but I played around with "make mrproper" and saw some
> > probably related errors.  Does this fix it?
> > 
> > ---8<---
> > 
> > Subject: [PATCH] stackvalidate: fix circular build dependencies
> > 
> > After "make mrproper" with CONFIG_STACK_VALIDATION enabled, I get the
> > following errors:
> > 
> >   make[2]: *** No rule to make target 'arch/x86/purgatory/purgatory.o', needed by 'arch/x86/purgatory/purgatory.ro'.  Stop.
> >   make[3]: *** No rule to make target 'scripts/mod/empty.o', needed by 'scripts/mod/elfconfig.h'.  Stop.
> > 
> > These are caused by circular dependencies.  The %.o pattern rules in
> > scripts/Makefile.build have the stackvalidate binary listed as a
> > dependency.  But stackvalidate gets built *after* archprepare and
> > scripts/mod, both of which build objects using the %.o pattern rules.
> > 
> > The STACKVALIDATE and STACKVALIDATE_foo.o variables are already used to
> > determine whether to validate a given object.  Also use them to
> > determine whether to create the pattern rule dependency.
> > 
> > Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com>
> > ---
> >  scripts/Makefile.build | 7 +++++--
> >  scripts/mod/Makefile   | 2 ++
> >  2 files changed, 7 insertions(+), 2 deletions(-)
> > 
> > diff --git a/scripts/Makefile.build b/scripts/Makefile.build
> > index a1270d3..ec96c51 100644
> > --- a/scripts/Makefile.build
> > +++ b/scripts/Makefile.build
> > @@ -243,7 +243,7 @@ endif
> >  
> >  ifdef CONFIG_STACK_VALIDATION
> >  
> > -stackvalidate = $(objtree)/scripts/stackvalidate/stackvalidate
> > +__stackvalidate = $(objtree)/scripts/stackvalidate/stackvalidate
> >  
> >  ifndef CONFIG_FRAME_POINTER
> >  nofp = --no-frame-pointer
> > @@ -251,9 +251,12 @@ endif
> >  
> >  # Set STACKVALIDATE_foo.o=n to skip stack validation for a file.
> >  # Set STACKVALIDATE=n to skip stack validation for a directory.
> > +stackvalidate = $(if $(patsubst n%,, \
> > +	$(STACKVALIDATE_$(basetarget).o)$(STACKVALIDATE)y), \
> > +	$(__stackvalidate))
> >  cmd_stackvalidate = $(if $(patsubst n%,, \
> >  	$(STACKVALIDATE_$(basetarget).o)$(STACKVALIDATE)y), \
> > -	$(stackvalidate) $(nofp) "$(@)";)
> > +	$(__stackvalidate) $(nofp) "$(@)";)
> >  
> >  endif # CONFIG_STACK_VALIDATION
> >  
> > diff --git a/scripts/mod/Makefile b/scripts/mod/Makefile
> > index c11212f..374c413 100644
> > --- a/scripts/mod/Makefile
> > +++ b/scripts/mod/Makefile
> > @@ -1,3 +1,5 @@
> > +STACKVALIDATE	:= n
> > +
> >  hostprogs-y	:= modpost mk_elfconfig
> >  always		:= $(hostprogs-y) empty.o
> >  
> > 
> 
> Josh,
> 
> I still get build failures and I've pared it down to x86_64 defconfig plus:
> CONFIG_MODVERSIONS=y
> CONFIG_STACK_VALIDATION=y

Are you still seeing the same error as before?  I think the errors you
saw were caused by stackvalidate choking on scripts/mod/empty.o.  But I
don't see how that's still possible with the above patch since I set
STACKVALIDATE to 'n' in scripts/mod/Makefile which should tell it to
skip the directory.

> And it seems like some modules may get the .tmp_foo.o treatment while
> others end up foo.o so something like the following will not work:
> 
> cmd_stackvalidate = $(if $(patsubst n%,, \
>         $(STACKVALIDATE_$(basetarget).o)$(STACKVALIDATE)y), \
>         $(__stackvalidate) $(nofp) "$(@D)/.tmp_$(@F)";)
> 
> In addition, I'm not sure if skipping modules like STACKVALIDATE_foo.o=n
> will still function properly for modversioned modules.

I'm not sure that's really a problem, since I don't think we'll need to
skip validation of any module code.  It's only used for skipping code
which runs outside of the kernel's normal operation (e.g., boot, vdso,
kexec).

-- 
Josh