From: Uwe Kleine-König u.kleine-koenig@pengutronix.de
stable inclusion from stable-v5.10.110 commit df2dc4cf71fa9bfc25115e69c41be97890aa1ece bugzilla: https://gitee.com/openeuler/kernel/issues/I574AL
Reference: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?id=...
--------------------------------
[ Upstream commit dedab69fd650ea74710b2e626e63fd35584ef773 ]
Set em485->active_timer = NULL isn't always enough to take out the stop timer. While there is a check that it acts in the right state (i.e. waiting for RTS-after-send to pass after sending some chars) but the following might happen:
- CPU1: some chars send, shifter becomes empty, stop tx timer armed - CPU0: more chars send before RTS-after-send expired - CPU0: shifter empty irq, port lock taken - CPU1: tx timer triggers, waits for port lock - CPU0: em485->active_timer = &em485->stop_tx_timer, hrtimer_start(), releases lock() - CPU1: get lock, see em485->active_timer == &em485->stop_tx_timer, tear down RTS too early
This fix bases on research done by Steffen Trumtrar.
Fixes: b86f86e8e7c5 ("serial: 8250: fix potential deadlock in rs485-mode") Signed-off-by: Uwe Kleine-König u.kleine-koenig@pengutronix.de Link: https://lore.kernel.org/r/20220215160236.344236-1-u.kleine-koenig@pengutroni... Signed-off-by: Greg Kroah-Hartman gregkh@linuxfoundation.org Signed-off-by: Sasha Levin sashal@kernel.org Signed-off-by: Yu Liao liaoyu15@huawei.com Reviewed-by: Wei Li liwei391@huawei.com Signed-off-by: Zheng Zengkai zhengzengkai@huawei.com --- drivers/tty/serial/8250/8250_port.c | 12 ++++++++++++ 1 file changed, 12 insertions(+)
diff --git a/drivers/tty/serial/8250/8250_port.c b/drivers/tty/serial/8250/8250_port.c index 7c07ebb37b1b..1733f03a7da7 100644 --- a/drivers/tty/serial/8250/8250_port.c +++ b/drivers/tty/serial/8250/8250_port.c @@ -1620,6 +1620,18 @@ static inline void start_tx_rs485(struct uart_port *port) struct uart_8250_port *up = up_to_u8250p(port); struct uart_8250_em485 *em485 = up->em485;
+ /* + * While serial8250_em485_handle_stop_tx() is a noop if + * em485->active_timer != &em485->stop_tx_timer, it might happen that + * the timer is still armed and triggers only after the current bunch of + * chars is send and em485->active_timer == &em485->stop_tx_timer again. + * So cancel the timer. There is still a theoretical race condition if + * the timer is already running and only comes around to check for + * em485->active_timer when &em485->stop_tx_timer is armed again. + */ + if (em485->active_timer == &em485->stop_tx_timer) + hrtimer_try_to_cancel(&em485->stop_tx_timer); + em485->active_timer = NULL;
if (em485->tx_stopped) {