Re: [RFC] namei: prevent sgid-hardlinks for unmapped gids

From: Kees Cook <keescook@chromium.org>
To: Dave Chinner <david@fromorbit.com>
Cc: Jan Kara <jack@suse.cz>, "Theodore Ts'o" <tytso@mit.edu>,
	Andy Lutomirski <luto@amacapital.net>,
	Theodore Tso <tytso@google.com>, Willy Tarreau <w@1wt.eu>,
	Dirk Steinmetz <public@rsjtdrjgfuzkfg.com>,
	Michael Kerrisk-manpages <mtk.manpages@gmail.com>,
	Serge Hallyn <serge.hallyn@ubuntu.com>,
	Seth Forshee <seth.forshee@canonical.com>,
	Alexander Viro <viro@zeniv.linux.org.uk>,
	Linux FS Devel <linux-fsdevel@vger.kernel.org>,
	LKML <linux-kernel@vger.kernel.org>,
	"Eric W . Biederman" <ebiederm@xmission.com>,
	Serge Hallyn <serge.hallyn@canonical.com>,
	"security@kernel.org" <security@kernel.org>
Subject: Re: [RFC] namei: prevent sgid-hardlinks for unmapped gids
Date: Thu, 19 Nov 2015 16:11:03 -0800	[thread overview]
Message-ID: <CAGXu5jKBN4E8uy0=_dYyddiZ8ogxS3pO2vF2nuFW57Lmo_x1MA@mail.gmail.com> (raw)
In-Reply-To: <20151119220241.GD19199@dastard>

On Thu, Nov 19, 2015 at 2:02 PM, Dave Chinner <david@fromorbit.com> wrote:
> On Thu, Nov 19, 2015 at 12:11:11PM -0800, Kees Cook wrote:
>> On Tue, Nov 10, 2015 at 7:08 AM, Jan Kara <jack@suse.cz> wrote:
>> > On Sat 07-11-15 21:02:06, Ted Tso wrote:
>> >> On Fri, Nov 06, 2015 at 09:05:57PM -0800, Kees Cook wrote:
>> >> > >>>> They're certainly not used early enough -- we need to remove suid when
>> >> > >>>> the page becomes writable via mmap (wp_page_shared), not when
>> >> > >>>> writeback happens, or at least not only when writeback happens.
>> >> > >>>
>> >> > >>> Well, I'm shy about the change there. For example, we don't strip in
>> >> > >>> on open(RDWR), just on write().
>> >> > >>
>> >> > >> I take it back. Hooking wp_page_shared looks expensive. :) Maybe we do
>> >> > >> need to hook the mmap?
>> >> > >
>> >> > > But file_update_time already pokes at the same (or nearby) cachelines,
>> >> > > I think -- why would it be expensive?  The whole thing could be
>> >> > > guarded by if (unlikely(is setuid)), right?
>> >> >
>> >> > Yeah, true. I added file_remove_privs calls near all the
>> >> > file_update_time calls, to no effect. Added to wp_page_shared too,
>> >> > nothing. Hmmm.
>> >>
>> >> Why not put the the should_remove_suid() call in
>> >> filemap_page_mkwrite(), or maybe do_page_mkwrite()?
>> >
>> > page_mkwrite() callbacks are IMHO the right place for this check (and
>> > change).  Just next to file_update_time() call. You get proper filesystem
>>
>> Should file_update_time() just be modified to include
>> file_remove_privs()? They seem to regularly go together.
>
> They might have similar call sites, but they are completely
> different operations. timestamp updates are optional, highly
> configurable and behaviour is filesystem implementation specific,
> whilst file_remove_privs() is mandatory and must be done in a
> crash-safe manner (i.e. via transactions). Hence, IMO, they need to
> be kept separate even if the call sites are similar.

Yeah, that was my worry too.

Okay, I think I've got it now. I had misunderstood the purpose of the
page_mkwrite variable in wp_page_reuse. My tests pass now. Patch on
it's way...

-Kees

>
> Cheers,
>
> Dave.
> --
> Dave Chinner
> david@fromorbit.com

-- 
Kees Cook
Chrome OS Security