serial: 8250: fix potential deadlock in rs485-mode

Canceling hrtimer when holding uart spinlock can deadlock.

CPU0: syscall write
          -> get uart port spinlock
              -> write uart
                  -> start_tx_rs485
                      -> hrtimer_cancel
                          -> wait for hrtimer callback to finish

CPU1: hrtimer IRQ
          -> run hrtimer
              -> em485_handle_stop_tx
                  -> get uart port spinlock

CPU0 is waiting for the hrtimer callback to finish, but the hrtimer
callback running on CPU1 is waiting to get the uart port spinlock.

This deadlock can be avoided by not canceling the hrtimers in these paths.
Setting active_timer=NULL can be done without accessing hrtimer,
and that will effectively cancel operations that would otherwise have been
performed by the hrtimer callback.

Signed-off-by: Tomas Melin <tomas.melin@vaisala.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

diff --git a/drivers/tty/serial/8250/8250_port.c b/drivers/tty/serial/8250/8250_port.c
index c2fb92f..d0d171b 100644 (file)
--- a/drivers/tty/serial/8250/8250_port.c
+++ b/drivers/tty/serial/8250/8250_port.c
@@ -1516,7 +1516,6 @@ static inline void __stop_tx(struct uart_8250_port *p)
                        return;
 
                em485->active_timer = NULL;
-               hrtimer_cancel(&em485->start_tx_timer);
 
                __stop_tx_rs485(p);
        }
@@ -1580,8 +1579,6 @@ static inline void start_tx_rs485(struct uart_port *port)
                serial8250_stop_rx(&up->port);
 
        em485->active_timer = NULL;
-       if (hrtimer_is_queued(&em485->stop_tx_timer))
-               hrtimer_cancel(&em485->stop_tx_timer);
 
        mcr = serial8250_in_MCR(up);
        if (!!(up->port.rs485.flags & SER_RS485_RTS_ON_SEND) !=