Bluetooth: RFCOMM: Fix possible deadlock on socket shutdown/release

[ Upstream commit 812e92b824c1db16c9519f8624d48a9901a0d38f ]

Due to change to switch to use lock_sock inside rfcomm_sk_state_change
the socket shutdown/release procedure can cause a deadlock:

    rfcomm_sock_shutdown():
      lock_sock();
      __rfcomm_sock_close():
        rfcomm_dlc_close():
          __rfcomm_dlc_close():
            rfcomm_dlc_lock();
            rfcomm_sk_state_change():
              lock_sock();

To fix this when the call __rfcomm_sock_close is now done without
holding the lock_sock since rfcomm_dlc_lock exists to protect
the dlc data there is no need to use lock_sock in that code path.

Link: https://lore.kernel.org/all/CAD+dNTsbuU4w+Y_P7o+VEN7BYCAbZuwZx2+tH+OTzCdcZF82YA@mail.gmail.com/
Fixes: b7ce436a5d79 ("Bluetooth: switch to lock_sock in RFCOMM")
Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

diff --git a/net/bluetooth/rfcomm/sock.c b/net/bluetooth/rfcomm/sock.c
index 4bf4ea6..21e24da 100644 (file)
--- a/net/bluetooth/rfcomm/sock.c
+++ b/net/bluetooth/rfcomm/sock.c
@@ -902,7 +902,10 @@ static int rfcomm_sock_shutdown(struct socket *sock, int how)
        lock_sock(sk);
        if (!sk->sk_shutdown) {
                sk->sk_shutdown = SHUTDOWN_MASK;
+
+               release_sock(sk);
                __rfcomm_sock_close(sk);
+               lock_sock(sk);
 
                if (sock_flag(sk, SOCK_LINGER) && sk->sk_lingertime &&
                    !(current->flags & PF_EXITING))