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:   Wed, 7 Jun 2023 16:37:55 -0700
From:   Kees Cook <keescook@chromium.org>
To:     Christian Marangi <ansuelsmth@gmail.com>
Cc:     Alexander Viro <viro@zeniv.linux.org.uk>,
        Christian Brauner <brauner@kernel.org>,
        Eric Biederman <ebiederm@xmission.com>,
        Mark Brown <broonie@kernel.org>,
        Dave Martin <Dave.Martin@arm.com>,
        Catalin Marinas <catalin.marinas@arm.com>,
        linux-fsdevel@vger.kernel.org, linux-mm@kvack.org,
        linux-kernel@vger.kernel.org, stable@vger.kernel.org
Subject: Re: [PATCH] binfmt_elf: dynamically allocate note.data in
 parse_elf_properties
Message-ID: <202306071636.1C35171CC@keescook>
References: <20230607144227.8956-1-ansuelsmth@gmail.com>
 <202306071417.79F70AC@keescook>
 <6480f938.1c0a0220.17a3a.0e1e@mx.google.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <6480f938.1c0a0220.17a3a.0e1e@mx.google.com>
Precedence: bulk

On Wed, Jun 07, 2023 at 08:31:58PM +0200, Christian Marangi wrote:
> On Wed, Jun 07, 2023 at 02:19:51PM -0700, Kees Cook wrote:
> > On Wed, Jun 07, 2023 at 04:42:27PM +0200, Christian Marangi wrote:
> > > Dynamically allocate note.data in parse_elf_properties to fix
> > > compilation warning on some arch.
> > 
> > I'd rather avoid dynamic allocation as much as possible in the exec
> > path, but we can balance it against how much it may happen.
> >
> 
> I guess there isn't a good way to handle this other than static global
> variables and kmalloc. But check the arch question for additional info
> on the case.
> 
> > > On some arch note.data exceed the stack limit for a single function and
> > > this cause the following compilation warning:
> > > fs/binfmt_elf.c: In function 'parse_elf_properties.isra':
> > > fs/binfmt_elf.c:821:1: error: the frame size of 1040 bytes is larger than 1024 bytes [-Werror=frame-larger-than=]
> > >   821 | }
> > >       | ^
> > > cc1: all warnings being treated as errors
> > 
> > Which architectures see this warning?
> > 
> 
> This is funny. On OpenWRT we are enforcing WERROR and we had FRAME_WARN
> hardcoded to 1024. (the option is set to 2048 on 64bit arch)

Ah-ha. Okay, I was wondering how you got that. :)

> ARCH_USE_GNU_PROPERTY is set only on arm64 that have a FRAME_WARN set to
> 2048.
> 
> So this was triggered by building arm64 with FRAME_WARN set to 1024.
> 
> Now with the configuration of 2048 the stack warn is not triggered, but
> I wonder if it may happen to have a 32bit system with
> ARCH_USE_GNU_PROPERTY. That would effectively trigger the warning.
> 
> So this is effectively a patch that fix a currently not possible
> configuration, since:
> 
> !IS_ENABLED(CONFIG_ARCH_USE_GNU_PROPERTY) will result in node.data
> effectively never allocated by the compiler are the function will return
> 0 on everything that doesn't have CONFIG_ARCH_USE_GNU_PROPERTY.
> 
> > > Fix this by dynamically allocating the array.
> > > Update the sizeof of the union to the biggest element allocated.
> > 
> > How common are these notes? I assume they're very common; I see them
> > even in /bin/true:
> > 
> > $ readelf -lW /bin/true | grep PROP
> >   GNU_PROPERTY   0x000338 0x0000000000000338 0x0000000000000338 0x000030 0x000030 R   0x8
> > 
> > -- 
> 
> Is there a way to check if this kmalloc actually cause perf regression?

I don't have a good benchmark besides just an exec loop. But since this
isn't reachable in a regular config, I'd rather keep things how there
already are.

-Kees

-- 
Kees Cook