Re: WARNING in apparmor_secid_to_secctx

From: Dmitry Vyukov <dvyukov@google.com>
To: Tetsuo Handa <penguin-kernel@i-love.sakura.ne.jp>
Cc: Casey Schaufler <casey@schaufler-ca.com>,
	Paul Moore <paul@paul-moore.com>,
	Stephen Smalley <sds@tycho.nsa.gov>,
	syzbot <syzbot+21016130b0580a9de3b5@syzkaller.appspotmail.com>,
	tyhicks@canonical.com,
	John Johansen <john.johansen@canonical.com>,
	James Morris <jmorris@namei.org>,
	LKML <linux-kernel@vger.kernel.org>,
	linux-security-module@vger.kernel.org,
	Serge Hallyn <serge@hallyn.com>,
	syzkaller-bugs <syzkaller-bugs@googlegroups.com>,
	Jeffrey Vander Stoep <jeffv@google.com>,
	SELinux <selinux@tycho.nsa.gov>,
	Russell Coker <russell@coker.com.au>,
	Laurent Bigonville <bigon@debian.org>,
	syzkaller <syzkaller@googlegroups.com>
Subject: Re: WARNING in apparmor_secid_to_secctx
Date: Fri, 1 Feb 2019 11:09:42 +0100	[thread overview]
Message-ID: <CACT4Y+Ynyy-bMK4cxhgNoAaauFdjzExuELM6u0amksbR4wy7GQ@mail.gmail.com> (raw)
In-Reply-To: <05340d28-36c2-267e-d54e-416fddfba211@i-love.sakura.ne.jp>

On Thu, Jan 31, 2019 at 1:23 AM Tetsuo Handa
<penguin-kernel@i-love.sakura.ne.jp> wrote:
>
> On 2019/01/30 23:45, Dmitry Vyukov wrote:
> >> Dmitry, is it possible to update configs for linux-next.git , for
> >> we want to test a big change in LSM which will go to Linux 5.1 ?
> >>
> >> TOMOYO security module (CONFIG_SECURITY_TOMOYO=y) can now coexist with
> >> SELinux/Smack/AppArmor security modules, and SafeSetID security module
> >> (CONFIG_SECURITY_SAFESETID=y) was added. Testing with these modules also
> >> enabled might find something...
> >
> > Hi,
> >
> > syzbot configs/cmdline args are stored here:
> > https://github.com/google/syzkaller/tree/master/dashboard/config
> >
> > I've tried to update to the latest kernel, the diff is below.
> > Few questions:
> > 1. How are modules enabled now?
> > We pass security=selinux of security=smack on command line. What do we
> > need to pass now to enable several modules at the same time?
>
> Removing security= parameter from kernel boot command line will do it.
>
>   security/apparmor/lsm.c:        .flags = LSM_FLAG_LEGACY_MAJOR | LSM_FLAG_EXCLUSIVE,
>   security/selinux/hooks.c:       .flags = LSM_FLAG_LEGACY_MAJOR | LSM_FLAG_EXCLUSIVE,
>   security/smack/smack_lsm.c:     .flags = LSM_FLAG_LEGACY_MAJOR | LSM_FLAG_EXCLUSIVE,
>   security/tomoyo/tomoyo.c:       .flags = LSM_FLAG_LEGACY_MAJOR,
>   security/security.c:                    if ((major->flags & LSM_FLAG_LEGACY_MAJOR) &&
>
> But this means that, if same kernel config/cmdline are used between
> linux-next.git and linux.git (etc.), syzbot will need to choose from
>
>  (a) stop sharing kernel cmdline between linux-next.git and linux.git (etc.)
>
> or
>
>  (b) stop sharing kernel config between SELinux, Smack and AppArmor
>
> or
>
>  (c) start testing after the LSM changes went to linux.git as Linux 5.1-rc1
>
> . Is (a) or (b) possible? If this is a too much change, (c) will be OK.

Thanks for the explanations.

Here is the change that I've come up with:
https://github.com/google/syzkaller/commit/aa53be276dc84aa8b3825b3416542447ff82b41a

I've disabled CONFIG_SECURITY_TOMOYO_OMIT_USERSPACE_LOADER (it
actually looked like omitting a user-space loader that I don't have is
the right thing to do, but okay, it indeed does not with =y).

For now I just enabled TOMOYO and SAFESETID.
I see the problem with making both linux-next and upstream work. If we
use a single config and lsm= cmdline argument, then on upstream all
kernels will use the same module (they won't understand lsm=). But if
we add security= then it will take precedence over lsm= on linux-next,
so we won't get stacked modules.

Let's go with (c) because I don't want an additional long-term maintenance cost.
If I understand it correctly later we will need to replace:
security=selinux
security=smack
security=apparmor

with:
lsm=yama,safesetid,integrity,selinux,tomoyo
lsm=yama,safesetid,integrity,smack,tomoyo
lsm=yama,safesetid,integrity,tomoyo,apparmor