Received-SPF: softfail (google.com: domain of transitioning linux-kernel-owner@vger.kernel.org does not designate 23.128.96.19 as permitted sender) client-ip=23.128.96.19;
Date:   Wed, 8 Jun 2022 15:51:10 +0200
From:   Michal =?iso-8859-1?Q?Koutn=FD?= <mkoutny@suse.com>
To:     shisiyuan <shisiyuan19870131@gmail.com>
Cc:     tj@kernel.org, hannes@cmpxchg.org, lizefan@huawei.com,
        cgroups@vger.kernel.org, linux-kernel@vger.kernel.org,
        shisiyuan <shisiyuan@xiaomi.com>
Subject: Re: [PATCH] cgroup: handle cset multiidentity issue when migration
Message-ID: <20220608135110.GA19399@blackbody.suse.cz>
References: <1654187688-27411-1-git-send-email-shisiyuan@xiaomi.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <1654187688-27411-1-git-send-email-shisiyuan@xiaomi.com>
User-Agent: Mutt/1.10.1 (2018-07-13)
Precedence: bulk

Hello.

On Fri, Jun 03, 2022 at 12:34:48AM +0800, shisiyuan <shisiyuan19870131@gmail.com> wrote:
> Bug code flow:
> cset X's initial refcount is 1.
> 1. cgroup_attach_task()
> 2. [For thread1]
>    cgroup_migrate_add_src()
>    [For thread2]
>    cgroup_migrate_add_src()
>      cset X is thread2's src_cset , ref->2,
>      and its mg_preload_node is added to
>      mgctx->preloaded_src_csets.
> 3. cgroup_migrate_prepare_dst()
>    [For thread1]
>    find_css_set()
>      cset X is thread1's dst_cset, ref->3
>    put_css_set()
>      ref->2 because cset X's mg_preload_node is not
>      empty(already in mgctx->preloaded_src_csets).
>    [For thread2]
>    find_css_cset()
>      cset X is also thread2's dst_cset, ref->3
>      then drop src_cset, ref->1
> [cgroup_free] ref->0
> 4. cgroup_migrate_execute
>    [For thread1]
>    ref -> 0xc0000000(UAF)

I'm trying to understand when this happens.
You migrate a process with two threads while one of them exits?

This should be properly synchronized with cgroup_threadgroup_rwsem, so I
don't understand where does the [cgroup_free] between 3. and 4. come
from.

Do you have a reproducer for this?

Thanks,
Michal