linux-kernel - Re: [PATCH] string.h: Mark 34 functions with __must

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Wed, 9 Oct 2019 09:31:41 -0700
From:   Nick Desaulniers <ndesaulniers@...gle.com>
To:     Dan Carpenter <dan.carpenter@...cle.com>
Cc:     Rasmus Villemoes <linux@...musvillemoes.dk>,
        Markus Elfring <Markus.Elfring@....de>,
        kernel-janitors@...r.kernel.org,
        Alexander Shishkin <alexander.shishkin@...ux.intel.com>,
        Andrew Morton <akpm@...ux-foundation.org>,
        Andy Shevchenko <andriy.shevchenko@...ux.intel.com>,
        Joe Perches <joe@...ches.com>,
        Kees Cook <keescook@...omium.org>,
        Steven Rostedt <rostedt@...dmis.org>,
        LKML <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH] string.h: Mark 34 functions with __must_check

On Wed, Oct 9, 2019 at 7:30 AM Dan Carpenter <dan.carpenter@...cle.com> wrote:
>
> On Wed, Oct 09, 2019 at 04:21:20PM +0200, Rasmus Villemoes wrote:
> > On 09/10/2019 15.56, Dan Carpenter wrote:
> > > That's because glibc strlen is annotated with __attribute_pure__ which
> > > means it has no side effects.
> >
> > I know, except it has nothing to do with glibc headers. Just try the
> > same thing in the kernel. gcc itself knows this about __builtin_strlen()
> > etc. If anything, we could annotate some of our non-standard functions
> > (say, memchr_inv) with __pure - then we'd both get the Wunused-value in
> > the nonsense cases, and allow gcc to optimize or reorder the calls.
>
> Huh.  You're right.  GCC already knows.  So this patch is pointless like
> you say.

Is it? None of the functions in include/linux/string.h are currently
marked __pure today.  (Side note, I'm surprised that any function that
accepts a pointer could be considered pure. I could reassign pointed
to value without changing the pointers value. I can see strlen being
"pure" for string literals, but not for char[].  This is something
I'll play with more, I've already spotted one missed optimization in
LLVM: https://bugs.llvm.org/show_bug.cgi?id=43624).

I think it would be an interesting study to see how often functions
that have return codes are ok to not check vs aren't ok (in a large
production codebase like the Linux kernel), similar to how 97% of
cases fallthrough is unintentional (which to me sounds like maybe the
default behavior of the language is incorrect).
-- 
Thanks,
~Nick Desaulniers