Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Subject: Re: [PATCH 1/1] kernel/module.c: wakeup processes in module_wq on
 module unload
To:     Konstantin Khorenko <khorenko@virtuozzo.com>,
        Jessica Yu <jeyu@kernel.org>, Barret Rhoden <brho@google.com>
Cc:     Andrey Ryabinin <aryabinin@virtuozzo.com>,
        linux-kernel@vger.kernel.org, David Arcari <darcari@redhat.com>,
        Heiko Carstens <heiko.carstens@de.ibm.com>
References: <20191113092950.15556-1-khorenko@virtuozzo.com>
 <20191113092950.15556-2-khorenko@virtuozzo.com>
From:   Prarit Bhargava <prarit@redhat.com>
Message-ID: <1b047b83-e6cd-9c56-9c55-a2ef6a3e16f5@redhat.com>
Date:   Wed, 13 Nov 2019 08:51:09 -0500
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:60.0) Gecko/20100101
 Thunderbird/60.6.1
MIME-Version: 1.0
In-Reply-To: <20191113092950.15556-2-khorenko@virtuozzo.com>
Content-Language: en-US
Content-Type: text/plain; charset=WINDOWS-1252
Content-Transfer-Encoding: quoted-printable
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk


On 11/13/19 4:29 AM, Konstantin Khorenko wrote:
> Fix the race between load and unload a kernel module.
>=20
> sys_delete_module()
>  try_stop_module()
>   mod->state =3D _GOING
> =09=09=09=09=09add_unformed_module()
> =09=09=09=09=09 old =3D find_module_all()
> =09=09=09=09=09 (old->state =3D=3D _GOING =3D>
> =09=09=09=09=09  wait_event_interruptible())
>=20
> =09=09=09=09=09 During pre-condition
> =09=09=09=09=09 finished_loading() rets 0
> =09=09=09=09=09 schedule()
> =09=09=09=09=09 (never gets waken up later)
>  free_module()
>   mod->state =3D _UNFORMED
>    list_del_rcu(&mod->list)
>    (dels mod from "modules" list)
>=20
> return
>=20
> The race above leads to modprobe hanging forever on loading
> a module.
>=20
> Error paths on loading module call wake_up_all(&module_wq) after
> freeing module, so let's do the same on straight module unload.
>=20
> Fixes: 6e6de3dee51a ("kernel/module.c: Only return -EEXIST
> for modules that have finished loading")
>=20
> Signed-off-by: Konstantin Khorenko <khorenko@virtuozzo.com>
> ---
>  kernel/module.c | 2 ++
>  1 file changed, 2 insertions(+)
>=20
> diff --git a/kernel/module.c b/kernel/module.c
> index ff2d7359a418..cb09a5f37a5f 100644
> --- a/kernel/module.c
> +++ b/kernel/module.c
> @@ -1033,6 +1033,8 @@ SYSCALL_DEFINE2(delete_module, const char __user *,=
 name_user,
>  =09strlcpy(last_unloaded_module, mod->name, sizeof(last_unloaded_module)=
);
> =20
>  =09free_module(mod);
> +=09/* someone could wait for the module in add_unformed_module() */
> +=09wake_up_all(&module_wq);

free_module() acts on that *this* module so it makes sense to include the
wake_up_all(&module_wq) call in delete_module.  IIRC I did try a module loa=
d &
unload stress test but it looks like my testing wasn't hard enough :(

In any case,

Reviewed-by: Prarit Bhargava <prarit@redhat.com>

P.

>  =09return 0;
>  out:
>  =09mutex_unlock(&module_mutex);
>=20