Re: [PATCH] binfmt_elf: Support segments with 0 filesz and misaligned starts

From: Kees Cook <keescook@chromium.org>
To: Sebastian Ott <sebott@redhat.com>,
	"Eric W. Biederman" <ebiederm@xmission.com>,
	Pedro Falcato <pedro.falcato@gmail.com>
Cc: "Thomas Weißschuh" <linux@weissschuh.net>,
	"Alexander Viro" <viro@zeniv.linux.org.uk>,
	"Christian Brauner" <brauner@kernel.org>,
	"Mark Brown" <broonie@kernel.org>, "Willy Tarreau" <w@1wt.eu>,
	sam@gentoo.org, "Rich Felker" <dalias@libc.org>,
	linux-fsdevel@vger.kernel.org, linux-mm@kvack.org,
	linux-kernel@vger.kernel.org, stable@vger.kernel.org
Subject: Re: [PATCH] binfmt_elf: Support segments with 0 filesz and misaligned starts
Date: Mon, 25 Sep 2023 10:06:01 -0700	[thread overview]
Message-ID: <202309251001.C050864@keescook> (raw)
In-Reply-To: <84e974d3-ae0d-9eb5-49b2-3348b7dcd336@redhat.com>

On Mon, Sep 25, 2023 at 05:27:12PM +0200, Sebastian Ott wrote:
> On Mon, 25 Sep 2023, Eric W. Biederman wrote:
> > 
> > Implement a helper elf_load that wraps elf_map and performs all
> > of the necessary work to ensure that when "memsz > filesz"
> > the bytes described by "memsz > filesz" are zeroed.
> > 
> > Link: https://lkml.kernel.org/r/20230914-bss-alloc-v1-1-78de67d2c6dd@weissschuh.net
> > Reported-by: Sebastian Ott <sebott@redhat.com>
> > Reported-by: Thomas Weißschuh <linux@weissschuh.net>
> > Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
> > ---
> > fs/binfmt_elf.c | 111 +++++++++++++++++++++---------------------------
> > 1 file changed, 48 insertions(+), 63 deletions(-)
> > 
> > Can you please test this one?

Eric thanks for doing this refactoring! This does look similar to the
earlier attempt:
https://lore.kernel.org/lkml/20221106021657.1145519-1-pedro.falcato@gmail.com/
and it's a bit easier to review.

> That one did the trick! The arm box booted successful, ran the binaries
> that were used for the repo of this issue, and ran the nolibc compiled
> binaries from kselftests that initially triggered the loader issues.

Thanks for testing! I need to dig out the other "weird" binaries (like
the mentioned ppc32 case) and see if I can get those tested too.

Pedro, are you able to test ppc64le musl libc with this patch too?

-Kees

-- 
Kees Cook