serial: 8250: fix potential deadlock in rs485-mode
authorTomas Melin <tomas.melin@vaisala.com>
Fri, 27 Oct 2017 12:16:30 +0000 (15:16 +0300)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Sat, 22 May 2021 08:57:43 +0000 (10:57 +0200)
commitf7d0f2440978f7d5e72b6f86af01e49f6e545690
tree4b8f389fba99788b50f1f4b08e1f8bae19f5a042
parente7f7cf9b93bd87c6ff91d733923a5bccf35c16d9
serial: 8250: fix potential deadlock in rs485-mode

[ Upstream commit b86f86e8e7c5264bb8f5835d60f9ec840d9f5a7a ]

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>
Signed-off-by: Sasha Levin <sashal@kernel.org>
drivers/tty/serial/8250/8250_port.c