Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Mon, 29 Aug 2022 14:19:59 -0700
From:   Kees Cook <keescook@chromium.org>
To:     Rasmus Villemoes <linux@rasmusvillemoes.dk>
Cc:     Gwan-gyeong Mun <gwan-gyeong.mun@intel.com>,
        Andrzej Hajda <andrzej.hajda@intel.com>,
        "Gustavo A. R. Silva" <gustavoars@kernel.org>,
        Nick Desaulniers <ndesaulniers@google.com>,
        linux-hardening@vger.kernel.org,
        Daniel Latypov <dlatypov@google.com>,
        Vitor Massaru Iha <vitor@massaru.org>,
        linux-kernel@vger.kernel.org
Subject: Re: [PATCH] overflow: Allow mixed type arguments
Message-ID: <202208291415.01271EA1@keescook>
References: <20220829204729.3409270-1-keescook@chromium.org>
 <d325bf21-efa4-b3e9-a0d8-3662a9f7ab1a@rasmusvillemoes.dk>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <d325bf21-efa4-b3e9-a0d8-3662a9f7ab1a@rasmusvillemoes.dk>
Precedence: bulk

On Mon, Aug 29, 2022 at 11:14:56PM +0200, Rasmus Villemoes wrote:
> On 29/08/2022 22.47, Kees Cook wrote:
> > When the check_[op]_overflow() helpers were introduced, all arguments were
> > required to be the same type to make the fallback macros simpler. However,
> > once the fallback macros were removed[1], it is fine to allow mixed
> > types, which makes using the helpers much more useful, as they can be
> > used to test for type-based overflows (e.g. adding two large ints but
> > storing into a u8), as would be handy in the drm core[2].
> > 
> > Remove the restriction, and add additional self-tests that exercise some
> > of the mixed-type overflow cases.
> 
> Makes sense. I'm a little worried about the implications for -stable
> backports to kernels that can still be built with gcc < 5.1, but we
> can't let that dictate what is done in mainline. And even people
> building old kernels shouldn't be using ancient compilers.

Right. I hope this will remain a theoretical problem, but if it really
comes up, the -stable patch can get some explicit type size checking or
something...

> 
> >  
> > -#define DEFINE_TEST_ARRAY(t)			\
> > -	static const struct test_ ## t {	\
> > -		t a, b;				\
> > -		t sum, diff, prod;		\
> > -		bool s_of, d_of, p_of;		\
> > -	} t ## _tests[]
> > +#define DEFINE_TEST_ARRAY_TYPED(t1, t2, t)		\
> > +	static const struct test_ ## t1 ## t2 ## t {	\
> > +		t1 a;					\
> > +		t2 b;					\
> > +		t sum, diff, prod;			\
> > +		bool s_of, d_of, p_of;			\
> > +	} t1 ## t2 ## t ## _tests[]
> 
> Can I get you to throw in some extra _, because this...
> 
> > +DEFINE_TEST_FUNC_TYPED(u32u32int, int, "%d");
> 
> ...makes my eyes hurt a little. Maybe even make it u32_u32__int, so it's
> emphasized that the order is [src op src -> tgt] and not [tgt = src op src].

Sure! Everything I tried hurt my eyes, so I opted for fewest characters.
;)

-- 
Kees Cook