MIME-Version: 1.0
References: <YbHTKUjEejZCLyhX@elver.google.com> <20211209201616.GU614@gate.crashing.org>
 <CANpmjNN4OAA_DM_KNLGJah3fk-PaZktGjziiu8ztf6fevZy5ug@mail.gmail.com>
In-Reply-To: <CANpmjNN4OAA_DM_KNLGJah3fk-PaZktGjziiu8ztf6fevZy5ug@mail.gmail.com>
From:   Marco Elver <elver@google.com>
Date:   Mon, 20 Dec 2021 08:00:00 +0100
Message-ID: <CANpmjNM3eSd9sxi-1tV0cRthJ0hudrME8nYdhYP=ttcWDoPNfg@mail.gmail.com>
Subject: Re: randomize_kstack: To init or not to init?
To:     Segher Boessenkool <segher@kernel.crashing.org>
Cc:     Kees Cook <keescook@chromium.org>,
        Thomas Gleixner <tglx@linutronix.de>,
        Nathan Chancellor <nathan@kernel.org>,
        Nick Desaulniers <ndesaulniers@google.com>,
        Elena Reshetova <elena.reshetova@intel.com>,
        Mark Rutland <mark.rutland@arm.com>,
        Peter Zijlstra <peterz@infradead.org>,
        Alexander Potapenko <glider@google.com>,
        Jann Horn <jannh@google.com>,
        Peter Collingbourne <pcc@google.com>,
        kasan-dev@googlegroups.com, linux-kernel@vger.kernel.org,
        llvm@lists.linux.dev, linux-toolchains@vger.kernel.org
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Thu, 9 Dec 2021 at 21:33, Marco Elver <elver@google.com> wrote:
> On Thu, 9 Dec 2021 at 21:19, Segher Boessenkool <segher@kernel.crashing.org> wrote:
> > On Thu, Dec 09, 2021 at 10:58:01AM +0100, Marco Elver wrote:
> > > Clang supports CONFIG_INIT_STACK_ALL_ZERO, which appears to be the
> > > default since dcb7c0b9461c2, which is why this came on my radar. And
> > > Clang also performs auto-init of allocas when auto-init is on
> > > (https://reviews.llvm.org/D60548), with no way to skip. As far as I'm
> > > aware, GCC 12's upcoming -ftrivial-auto-var-init= doesn't yet auto-init
> > > allocas.
> >
> > The space allocated by alloca is not an automatic variable, so of course
> > it is not affected by this compiler flag.  And it should not, this flag
> > is explicitly for *small fixed-size* stack variables (initialising
> > others can be much too expensive).
> >
> > >       C. Introduce a new __builtin_alloca_uninitialized().
> >
> > That is completely backwards.  That is the normal behaviour of alloca
> > already.  Also you can get __builtin_alloca inserted by the compiler
> > (for a variable length array for example), and you typically do not want
> > those initialised either, for the same reasons.
>
> You're right, if we're strict about it, initializing allocas is
> technically out-of-scope of that feature.
>
> So, option D: Add a param to control this, and probably it shouldn't
> do it by default. Let's see how far that gets then.

Just an update: after some discussion, the Clang side says that
alloca() is in scope, because the intent is that trivially initialized
"automatic stack storage" should be handled by ftrivial-auto-var-init.
And alloca() is automatic stack storage:
https://www.gnu.org/software/libc/manual/html_node/Variable-Size-Automatic.html

So currently it looks like the builtin is the only solution in this case.

Thanks.