Received-SPF: pass (google.com: domain of linux-bluetooth-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Tue, 10 Jan 2023 17:03:57 -0800
From:   Saeed Mahameed <saeed@kernel.org>
To:     Leon Romanovsky <leon@kernel.org>
Cc:     Ying Hsu <yinghsu@chromium.org>, linux-bluetooth@vger.kernel.org,
        chromeos-bluetooth-upstreaming@chromium.org,
        "David S. Miller" <davem@davemloft.net>,
        Eric Dumazet <edumazet@google.com>,
        Jakub Kicinski <kuba@kernel.org>,
        Johan Hedberg <johan.hedberg@gmail.com>,
        Luiz Augusto von Dentz <luiz.dentz@gmail.com>,
        Marcel Holtmann <marcel@holtmann.org>,
        Paolo Abeni <pabeni@redhat.com>, linux-kernel@vger.kernel.org,
        netdev@vger.kernel.org
Subject: Re: [PATCH] Bluetooth: Fix possible deadlock in
 rfcomm_sk_state_change
Message-ID: <Y74K/VKfB8K691OZ@x130>
References: <20230103111221.1.I1f29bb547a03e9adfe2e6754212f9d14a2e39c4b@changeid>
 <Y7UiDn3Gi5YdNIoC@unreal>
 <Y7dqKnQe8UUeQ/CD@x130>
 <Y7qXIZNsju8dCzqu@unreal>
 <Y70q4ZsdQNP9GIbF@x130>
 <Y71Yl8obzOWO/c+p@unreal>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii; format=flowed
Content-Disposition: inline
In-Reply-To: <Y71Yl8obzOWO/c+p@unreal>
Precedence: bulk

On 10 Jan 14:22, Leon Romanovsky wrote:
>On Tue, Jan 10, 2023 at 01:07:45AM -0800, Saeed Mahameed wrote:
>> On 08 Jan 12:12, Leon Romanovsky wrote:
>> > On Thu, Jan 05, 2023 at 04:24:10PM -0800, Saeed Mahameed wrote:
>> > > On 04 Jan 08:51, Leon Romanovsky wrote:
>> > > > On Tue, Jan 03, 2023 at 11:12:46AM +0000, Ying Hsu wrote:
>> > > > > There's a possible deadlock when two processes are connecting
>> > > > > and closing concurrently:
>> > > > >   + CPU0: __rfcomm_dlc_close locks rfcomm and then calls
>> > > > >   rfcomm_sk_state_change which locks the sock.
>> > > > >   + CPU1: rfcomm_sock_connect locks the socket and then calls
>> > > > >   rfcomm_dlc_open which locks rfcomm.
>> > > > >
>> > > > > Here's the call trace:
>> > > > >
>> > > > > -> #2 (&d->lock){+.+.}-{3:3}:
>> > > > >        __mutex_lock_common kernel/locking/mutex.c:603 [inline]
>> > > > >        __mutex_lock0x12f/0x1360 kernel/locking/mutex.c:747
>> > > > >        __rfcomm_dlc_close+0x15d/0x890 net/bluetooth/rfcomm/core.c:487
>> > > > >        rfcomm_dlc_close+1e9/0x240 net/bluetooth/rfcomm/core.c:520
>> > > > >        __rfcomm_sock_close+0x13c/0x250 net/bluetooth/rfcomm/sock.c:220
>> > > > >        rfcomm_sock_shutdown+0xd8/0x230 net/bluetooth/rfcomm/sock.c:907
>> > > > >        rfcomm_sock_release+0x68/0x140 net/bluetooth/rfcomm/sock.c:928
>> > > > >        __sock_release+0xcd/0x280 net/socket.c:650
>> > > > >        sock_close+0x1c/0x20 net/socket.c:1365
>> > > > >        __fput+0x27c/0xa90 fs/file_table.c:320
>> > > > >        task_work_run+0x16f/0x270 kernel/task_work.c:179
>> > > > >        exit_task_work include/linux/task_work.h:38 [inline]
>> > > > >        do_exit+0xaa8/0x2950 kernel/exit.c:867
>> > > > >        do_group_exit+0xd4/0x2a0 kernel/exit.c:1012
>> > > > >        get_signal+0x21c3/0x2450 kernel/signal.c:2859
>> > > > >        arch_do_signal_or_restart+0x79/0x5c0 arch/x86/kernel/signal.c:306
>> > > > >        exit_to_user_mode_loop kernel/entry/common.c:168 [inline]
>> > > > >        exit_to_user_mode_prepare+0x15f/0x250 kernel/entry/common.c:203
>> > > > >        __syscall_exit_to_user_mode_work kernel/entry/common.c:285 [inline]
>> > > > >        syscall_exit_to_user_mode+0x1d/0x50 kernel/entry/common.c:296
>> > > > >        do_syscall_64+0x46/0xb0 arch/x86/entry/common.c:86
>> > > > >        entry_SYSCALL_64_after_hwframe+0x63/0xcd
>> > > > >
>> > > > > -> #1 (rfcomm_mutex){+.+.}-{3:3}:
>> > > > >        __mutex_lock_common kernel/locking/mutex.c:603 [inline]
>> > > > >        __mutex_lock+0x12f/0x1360 kernel/locking/mutex.c:747
>> > > > >        rfcomm_dlc_open+0x93/0xa80 net/bluetooth/rfcomm/core.c:425
>> > > > >        rfcomm_sock_connect+0x329/0x450 net/bluetooth/rfcomm/sock.c:413
>> > > > >        __sys_connect_file+0x153/0x1a0 net/socket.c:1976
>> > > > >        __sys_connect+0x165/0x1a0 net/socket.c:1993
>> > > > >        __do_sys_connect net/socket.c:2003 [inline]
>> > > > >        __se_sys_connect net/socket.c:2000 [inline]
>> > > > >        __x64_sys_connect+0x73/0xb0 net/socket.c:2000
>> > > > >        do_syscall_x64 arch/x86/entry/common.c:50 [inline]
>> > > > >        do_syscall_64+0x39/0xb0 arch/x86/entry/common.c:80
>> > > > >        entry_SYSCALL_64_after_hwframe+0x63/0xcd
>> > > > >
>> > > > > -> #0 (sk_lock-AF_BLUETOOTH-BTPROTO_RFCOMM){+.+.}-{0:0}:
>> > > > >        check_prev_add kernel/locking/lockdep.c:3097 [inline]
>> > > > >        check_prevs_add kernel/locking/lockdep.c:3216 [inline]
>> > > > >        validate_chain kernel/locking/lockdep.c:3831 [inline]
>> > > > >        __lock_acquire+0x2a43/0x56d0 kernel/locking/lockdep.c:5055
>> > > > >        lock_acquire kernel/locking/lockdep.c:5668 [inline]
>> > > > >        lock_acquire+0x1e3/0x630 kernel/locking/lockdep.c:5633
>> > > > >        lock_sock_nested+0x3a/0xf0 net/core/sock.c:3470
>> > > > >        lock_sock include/net/sock.h:1725 [inline]
>> > > > >        rfcomm_sk_state_change+0x6d/0x3a0 net/bluetooth/rfcomm/sock.c:73
>> > > > >        __rfcomm_dlc_close+0x1b1/0x890 net/bluetooth/rfcomm/core.c:489
>> > > > >        rfcomm_dlc_close+0x1e9/0x240 net/bluetooth/rfcomm/core.c:520
>> > > > >        __rfcomm_sock_close+0x13c/0x250 net/bluetooth/rfcomm/sock.c:220
>> > > > >        rfcomm_sock_shutdown+0xd8/0x230 net/bluetooth/rfcomm/sock.c:907
>> > > > >        rfcomm_sock_release+0x68/0x140 net/bluetooth/rfcomm/sock.c:928
>> > > > >        __sock_release+0xcd/0x280 net/socket.c:650
>> > > > >        sock_close+0x1c/0x20 net/socket.c:1365
>> > > > >        __fput+0x27c/0xa90 fs/file_table.c:320
>> > > > >        task_work_run+0x16f/0x270 kernel/task_work.c:179
>> > > > >        exit_task_work include/linux/task_work.h:38 [inline]
>> > > > >        do_exit+0xaa8/0x2950 kernel/exit.c:867
>> > > > >        do_group_exit+0xd4/0x2a0 kernel/exit.c:1012
>> > > > >        get_signal+0x21c3/0x2450 kernel/signal.c:2859
>> > > > >        arch_do_signal_or_restart+0x79/0x5c0 arch/x86/kernel/signal.c:306
>> > > > >        exit_to_user_mode_loop kernel/entry/common.c:168 [inline]
>> > > > >        exit_to_user_mode_prepare+0x15f/0x250 kernel/entry/common.c:203
>> > > > >        __syscall_exit_to_user_mode_work kernel/entry/common.c:285 [inline]
>> > > > >        syscall_exit_to_user_mode+0x1d/0x50 kernel/entry/common.c:296
>> > > > >        do_syscall_64+0x46/0xb0 arch/x86/entry/common.c:86
>> > > > >        entry_SYSCALL_64_after_hwframe+0x63/0xcd
>> > > > >
>> > > > > Signed-off-by: Ying Hsu <yinghsu@chromium.org>
>> > > > > ---
>> > > > > This commit has been tested with a C reproducer on qemu-x86_64.
>> > > > >
>> > > > >  net/bluetooth/rfcomm/sock.c | 2 ++
>> > > > >  1 file changed, 2 insertions(+)
>> > > > >
>> > > > > diff --git a/net/bluetooth/rfcomm/sock.c b/net/bluetooth/rfcomm/sock.c
>> > > > > index 21e24da4847f..29f9a88a3dc8 100644
>> > > > > --- a/net/bluetooth/rfcomm/sock.c
>> > > > > +++ b/net/bluetooth/rfcomm/sock.c
>> > > > > @@ -410,8 +410,10 @@ static int rfcomm_sock_connect(struct socket *sock, struct sockaddr *addr, int a
>> > > > >  	d->sec_level = rfcomm_pi(sk)->sec_level;
>> > > > >  	d->role_switch = rfcomm_pi(sk)->role_switch;
>> > > > >
>> > > > > +	release_sock(sk);
>> > > > >  	err = rfcomm_dlc_open(d, &rfcomm_pi(sk)->src, &sa->rc_bdaddr,
>> > > >                                           ^^^^
>> > > > Are you sure that "sk" still exists here after you called to release_sock(sk)?
>> > > > What prevents from use-after-free here?
>> > > >
>> > >
>> > > sk must be valid to be locked in first place.
>> >
>> > It is, but after it is released it won't.
>> >
>>
>> the code is symmetric: you hold the sk lock then do your thing and then
>> release it.
>>
>> if you claim that sk can be freed by another process after you released it,
>> then due to symmetry it also can be freed before you locked it, unless
>
>So we can extend your logic and say what the lock_sock() in the beginning of
>rfcomm_sock_connect() is not needed too.
>

No, it is needed to synchronize the sk state as Luiz explained. 

consider the following:
<--- point A
lock(object);
do_something(object);
unlock(object);
<--- point B

if object can be freed at point B then it's also subject to being freed
at point A, unless do_something(); releases an extra reference.. 
which i am sure is not the case here.

So a higher level synchronization (ref counting) is required to manage the
sk life cycle .. e.g sock_hold/sock_put, but please read the comment on
sock_hold, there are some specific conditions to guarantee sk validity when
using sock_hold()