diff --git a/drivers/char/amiserial.c b/drivers/char/amiserial.c
index 1b21a7adeb58..8228e61bb54d 100644
--- a/drivers/char/amiserial.c
+++ b/drivers/char/amiserial.c
@@ -1528,6 +1528,7 @@ static void rs_wait_until_sent(struct tty_struct *tty, int timeout)
 {
 	struct async_struct * info = tty->driver_data;
 	unsigned long orig_jiffies, char_time;
+	int tty_was_locked = tty_locked();
 	int lsr;
 
 	if (serial_paranoia_check(info, tty->name, "rs_wait_until_sent"))
@@ -1538,7 +1539,12 @@ static void rs_wait_until_sent(struct tty_struct *tty, int timeout)
 
 	orig_jiffies = jiffies;
 
-	tty_lock_nested(); /* tty_wait_until_sent is called from lots of places */
+	/*
+	 * tty_wait_until_sent is called from lots of places,
+	 * with or without the BTM.
+	 */
+	if (!tty_was_locked)
+		tty_lock();
 	/*
 	 * Set the check interval to be 1/5 of the estimated time to
 	 * send a single character, and make it at least 1.  The check
@@ -1579,7 +1585,8 @@ static void rs_wait_until_sent(struct tty_struct *tty, int timeout)
 			break;
 	}
 	__set_current_state(TASK_RUNNING);
-	tty_unlock();
+	if (!tty_was_locked)
+		tty_unlock();
 #ifdef SERIAL_DEBUG_RS_WAIT_UNTIL_SENT
 	printk("lsr = %d (jiff=%lu)...done\n", lsr, jiffies);
 #endif
diff --git a/drivers/serial/68360serial.c b/drivers/serial/68360serial.c
index 16f5f2fab032..edcf1cc089d1 100644
--- a/drivers/serial/68360serial.c
+++ b/drivers/serial/68360serial.c
@@ -1705,7 +1705,6 @@ static void rs_360_wait_until_sent(struct tty_struct *tty, int timeout)
 	printk("jiff=%lu...", jiffies);
 #endif
 
-	tty_lock_nested(); /* always held already since we come from ->close */
 	/* We go through the loop at least once because we can't tell
 	 * exactly when the last character exits the shifter.  There can
 	 * be at least two characters waiting to be sent after the buffers
@@ -1734,7 +1733,6 @@ static void rs_360_wait_until_sent(struct tty_struct *tty, int timeout)
 			bdp--;
 	} while (bdp->status & BD_SC_READY);
 	current->state = TASK_RUNNING;
-	tty_unlock();
 #ifdef SERIAL_DEBUG_RS_WAIT_UNTIL_SENT
 	printk("lsr = %d (jiff=%lu)...done\n", lsr, jiffies);
 #endif
diff --git a/drivers/serial/crisv10.c b/drivers/serial/crisv10.c
index 94bfb9f238e1..8e356c5eeae0 100644
--- a/drivers/serial/crisv10.c
+++ b/drivers/serial/crisv10.c
@@ -3935,7 +3935,6 @@ static void rs_wait_until_sent(struct tty_struct *tty, int timeout)
 	 * Check R_DMA_CHx_STATUS bit 0-6=number of available bytes in FIFO
 	 * R_DMA_CHx_HWSW bit 31-16=nbr of bytes left in DMA buffer (0=64k)
 	 */
-	tty_lock_nested(); /* locked already when coming from close */
 	orig_jiffies = jiffies;
 	while (info->xmit.head != info->xmit.tail || /* More in send queue */
 	       (*info->ostatusadr & 0x007f) ||  /* more in FIFO */
@@ -3952,7 +3951,6 @@ static void rs_wait_until_sent(struct tty_struct *tty, int timeout)
 			curr_time_usec - info->last_tx_active_usec;
 	}
 	set_current_state(TASK_RUNNING);
-	tty_unlock();
 }
 
 /*
diff --git a/drivers/serial/serial_core.c b/drivers/serial/serial_core.c
index 851d7c29132b..cd8511298bcb 100644
--- a/drivers/serial/serial_core.c
+++ b/drivers/serial/serial_core.c
@@ -60,7 +60,7 @@ static struct lock_class_key port_lock_key;
 
 static void uart_change_speed(struct tty_struct *tty, struct uart_state *state,
 					struct ktermios *old_termios);
-static void uart_wait_until_sent(struct tty_struct *tty, int timeout);
+static void __uart_wait_until_sent(struct uart_port *port, int timeout);
 static void uart_change_pm(struct uart_state *state, int pm_state);
 
 /*
@@ -1322,8 +1322,16 @@ static void uart_close(struct tty_struct *tty, struct file *filp)
 	tty->closing = 1;
 	spin_unlock_irqrestore(&port->lock, flags);
 
-	if (port->closing_wait != ASYNC_CLOSING_WAIT_NONE)
-		tty_wait_until_sent(tty, msecs_to_jiffies(port->closing_wait));
+	if (port->closing_wait != ASYNC_CLOSING_WAIT_NONE) {
+		/*
+		 * hack: open-coded tty_wait_until_sent to avoid
+		 * recursive tty_lock
+		 */
+		long timeout = msecs_to_jiffies(port->closing_wait);
+		if (wait_event_interruptible_timeout(tty->write_wait,
+				!tty_chars_in_buffer(tty), timeout) >= 0)
+			__uart_wait_until_sent(uport, timeout);
+	}
 
 	/*
 	 * At this point, we stop accepting input.  To do this, we
@@ -1339,7 +1347,7 @@ static void uart_close(struct tty_struct *tty, struct file *filp)
 		 * has completely drained; this is especially
 		 * important if there is a transmit FIFO!
 		 */
-		uart_wait_until_sent(tty, uport->timeout);
+		__uart_wait_until_sent(uport, uport->timeout);
 	}
 
 	uart_shutdown(tty, state);
@@ -1373,17 +1381,13 @@ done:
 	mutex_unlock(&port->mutex);
 }
 
-static void uart_wait_until_sent(struct tty_struct *tty, int timeout)
+static void __uart_wait_until_sent(struct uart_port *port, int timeout)
 {
-	struct uart_state *state = tty->driver_data;
-	struct uart_port *port = state->uart_port;
 	unsigned long char_time, expire;
 
 	if (port->type == PORT_UNKNOWN || port->fifosize == 0)
 		return;
 
-	tty_lock_nested(); /* already locked when coming from close */
-
 	/*
 	 * Set the check interval to be 1/5 of the estimated time to
 	 * send a single character, and make it at least 1.  The check
@@ -1429,6 +1433,15 @@ static void uart_wait_until_sent(struct tty_struct *tty, int timeout)
 			break;
 	}
 	set_current_state(TASK_RUNNING); /* might not be needed */
+}
+
+static void uart_wait_until_sent(struct tty_struct *tty, int timeout)
+{
+	struct uart_state *state = tty->driver_data;
+	struct uart_port *port = state->uart_port;
+
+	tty_lock();
+	__uart_wait_until_sent(port, timeout);
 	tty_unlock();
 }