From:   ebiederm@xmission.com (Eric W. Biederman)
To:     Michal =?utf-8?Q?Koutn=C3=BD?= <mkoutny@suse.com>
Cc:     linux-kernel@vger.kernel.org, Jens Axboe <axboe@kernel.dk>,
        Kees Cook <keescook@chromium.org>,
        Oleg Nesterov <oleg@redhat.com>,
        Peter Zijlstra <peterz@infradead.org>,
        Thomas Gleixner <tglx@linutronix.de>,
        Jim Newsome <jnewsome@torproject.org>,
        Alexey Gladkov <legion@kernel.org>, Tejun Heo <tj@kernel.org>,
        security@kernel.org, Andy Lutomirski <luto@amacapital.net>,
        Jann Horn <jannh@google.com>
References: <20211208180501.11969-1-mkoutny@suse.com>
        <87sfv3540t.fsf@email.froward.int.ebiederm.org>
        <YbPe4+K+RlDgccIh@blackbook>
Date:   Mon, 13 Dec 2021 09:24:55 -0600
In-Reply-To: <YbPe4+K+RlDgccIh@blackbook> ("Michal \=\?utf-8\?Q\?Koutn\=C3\=BD\?\=
 \=\?utf-8\?Q\?\=22's\?\= message of "Sat,
        11 Dec 2021 00:12:35 +0100")
Message-ID: <874k7c34u0.fsf@email.froward.int.ebiederm.org>
User-Agent: Gnus/5.13 (Gnus v5.13) Emacs/26.1 (gnu/linux)
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 8BIT
Subject: Re: [PATCH] exit: Retain nsproxy for exit_task_work() work entries
Precedence: bulk

Michal Koutný <mkoutny@suse.com> writes:

> On Wed, Dec 08, 2021 at 12:45:54PM -0600, "Eric W. Biederman" <ebiederm@xmission.com> wrote:
>> TL;DR the cgroup file system is checking permissions at write time.
>
> Thank you for bringing that up (handled in a separate thread now).
>
>> I think I follow your reasoning and I think it will even fix the issue
>> but no.
>
> FTR, part of Tejun's series [1] ensures that cgroup_ns is accessed
> directly without nsproxy and a reference to it is kept while the file
> is opened. I.e. that'd properly fix this particular crash reported by
> syzbot.
>
>> Please don't apply this patch.
>> 
>> exit_task_work running after exit_task_namespaces is the messenger
>> that just told us about something ugly.
>
> In (my) theory some other task_work callbacks could (transitively) rely
> on the current->nsproxy which could still be cleared by
> exit_task_namespaces().
> Is there another reason why to have exit_task_namespaces() before
> exit_task_work()?

We already have the principle that things are going to be cleaned up
before exit_task_work is called and exit_files depends upon that.

So I think the burden is to find a good reason why exit_task_work should
move not to defend it.

If we don't want things cleaned up before exit_task_work it should come
at the start of do_exit and exit_files and others need to stop depending
upon it.  Which seems like challenging change to make.

Eric