[PATCH v3 0/4] serial: fix ioctl hangup race

[PATCH v3 0/4] serial: fix ioctl hangup race

[Johan Hovold]

The tty ioctls can race with hangup and end up calling into a tty
driver for a device that is already gone or powered down.

Included in v3 are also two related fixes for issues highlighted by
Sashiko.

Johan


Changes in v3
 - revert guards in uart_wait_modem_status()
 - fix race in TIOCMIWAIT (new)
 - abort TIOCMIWAIT on hangup (new)

Changes in v2
 - include TIOCMIWAIT which also access hardware
  - replace broken scoped_guard() construct in wait loop


Johan Hovold (4):
  serial: revert guards in uart_wait_modem_status()
  serial: fix ioctl hangup race
  serial: fix TIOCMIWAIT race
  serial: abort TIOCMIWAIT on hangup

 drivers/tty/serial/serial_core.c | 50 +++++++++++++++++++++++---------
 1 file changed, 37 insertions(+), 13 deletions(-)

-- 
2.55.0

[PATCH v3 1/4] serial: revert guards in uart_wait_modem_status()

[Johan Hovold]

Mixing scope-based and regular cleanup is discouraged and
uart_port_deref() used by uart_wait_modem_status() falls in the latter
category.

Revert the premature guard conversion in preparation for fixing a hangup
race.

Fixes: 56609c050051 ("serial: serial_core: use guard()s")
Cc: stable@vger.kernel.org	# 6.18
Cc: Jiri Slaby (SUSE) <jirislaby@kernel.org>
Signed-off-by: Johan Hovold <johan@kernel.org>
---
 drivers/tty/serial/serial_core.c | 13 +++++++------
 1 file changed, 7 insertions(+), 6 deletions(-)

diff --git a/drivers/tty/serial/serial_core.c b/drivers/tty/serial/serial_core.c
index 95774b0f1484..1553bc6cbe7b 100644
--- a/drivers/tty/serial/serial_core.c
+++ b/drivers/tty/serial/serial_core.c
@@ -1213,15 +1213,16 @@ static int uart_wait_modem_status(struct uart_state *state, unsigned long arg)
 	uport = uart_port_ref(state);
 	if (!uport)
 		return -EIO;
-	scoped_guard(uart_port_lock_irq, uport) {
-		memcpy(&cprev, &uport->icount, sizeof(struct uart_icount));
-		uart_enable_ms(uport);
-	}
+	uart_port_lock_irq(uport);
+	memcpy(&cprev, &uport->icount, sizeof(struct uart_icount));
+	uart_enable_ms(uport);
+	uart_port_unlock_irq(uport);
 
 	add_wait_queue(&port->delta_msr_wait, &wait);
 	for (;;) {
-		scoped_guard(uart_port_lock_irq, uport)
-			memcpy(&cnow, &uport->icount, sizeof(struct uart_icount));
+		uart_port_lock_irq(uport);
+		memcpy(&cnow, &uport->icount, sizeof(struct uart_icount));
+		uart_port_unlock_irq(uport);
 
 		set_current_state(TASK_INTERRUPTIBLE);
 
-- 
2.55.0

[PATCH v3 2/4] serial: fix ioctl hangup race

[Johan Hovold]

The tty ioctls can race with hangup and end up calling into a tty
driver for a device that is already gone or powered down.

Add the missing checks to make sure the port has not been hung up before
accessing the hardware to avoid issues like kernel panic due to
unclocked accesses.

Note that TIOCGSERIAL, TIOCGICOUNT do not access hardware and are
therefore not affected by the race.

Fixes: 04f378b198da ("tty: BKL pushdown")
Cc: stable@vger.kernel.org	# 2.6.26
Signed-off-by: Johan Hovold <johan@kernel.org>
---
 drivers/tty/serial/serial_core.c | 32 +++++++++++++++++++++++++-------
 1 file changed, 25 insertions(+), 7 deletions(-)

diff --git a/drivers/tty/serial/serial_core.c b/drivers/tty/serial/serial_core.c
index 1553bc6cbe7b..4213fa3dc988 100644
--- a/drivers/tty/serial/serial_core.c
+++ b/drivers/tty/serial/serial_core.c
@@ -896,7 +896,7 @@ static int uart_set_info(struct tty_struct *tty, struct tty_port *port,
 	upf_t old_flags, new_flags;
 	int retval;
 
-	if (!uport)
+	if (!uport || tty_io_error(tty))
 		return -EIO;
 
 	new_port = new_info->port;
@@ -1119,7 +1119,7 @@ static int uart_break_ctl(struct tty_struct *tty, int break_state)
 	guard(mutex)(&port->mutex);
 
 	uport = uart_port_check(state);
-	if (!uport)
+	if (!uport || tty_io_error(tty))
 		return -EIO;
 
 	if (uport->type != PORT_UNKNOWN && uport->ops->break_ctl)
@@ -1144,7 +1144,7 @@ static int uart_do_autoconfig(struct tty_struct *tty, struct uart_state *state)
 	 */
 	scoped_cond_guard(mutex_intr, return -ERESTARTSYS, &port->mutex) {
 		uport = uart_port_check(state);
-		if (!uport)
+		if (!uport || tty_io_error(tty))
 			return -EIO;
 
 		if (tty_port_users(port) != 1)
@@ -1199,7 +1199,7 @@ static void uart_enable_ms(struct uart_port *uport)
  * FIXME: This wants extracting into a common all driver implementation
  * of TIOCMWAIT using tty_port.
  */
-static int uart_wait_modem_status(struct uart_state *state, unsigned long arg)
+static int uart_wait_modem_status(struct tty_struct *tty, struct uart_state *state, unsigned long arg)
 {
 	struct uart_port *uport;
 	struct tty_port *port = &state->port;
@@ -1213,11 +1213,21 @@ static int uart_wait_modem_status(struct uart_state *state, unsigned long arg)
 	uport = uart_port_ref(state);
 	if (!uport)
 		return -EIO;
+
+	mutex_lock(&port->mutex);
+	if (tty_io_error(tty)) {
+		mutex_unlock(&port->mutex);
+		ret = -EIO;
+		goto out_deref;
+	}
+
 	uart_port_lock_irq(uport);
 	memcpy(&cprev, &uport->icount, sizeof(struct uart_icount));
 	uart_enable_ms(uport);
 	uart_port_unlock_irq(uport);
 
+	mutex_unlock(&port->mutex);
+
 	add_wait_queue(&port->delta_msr_wait, &wait);
 	for (;;) {
 		uart_port_lock_irq(uport);
@@ -1246,6 +1256,7 @@ static int uart_wait_modem_status(struct uart_state *state, unsigned long arg)
 	}
 	__set_current_state(TASK_RUNNING);
 	remove_wait_queue(&port->delta_msr_wait, &wait);
+out_deref:
 	uart_port_deref(uport);
 
 	return ret;
@@ -1568,7 +1579,7 @@ uart_ioctl(struct tty_struct *tty, unsigned int cmd, unsigned long arg)
 
 	/* This should only be used when the hardware is present. */
 	if (cmd == TIOCMIWAIT)
-		return uart_wait_modem_status(state, arg);
+		return uart_wait_modem_status(tty, state, arg);
 
 	/* rs485_config requires more locking than others */
 	if (cmd == TIOCSRS485)
@@ -1647,7 +1658,7 @@ static void uart_set_termios(struct tty_struct *tty,
 	guard(mutex)(&state->port.mutex);
 
 	uport = uart_port_check(state);
-	if (!uport)
+	if (!uport || tty_io_error(tty))
 		return;
 
 	/*
@@ -1799,7 +1810,14 @@ static void uart_wait_until_sent(struct tty_struct *tty, int timeout)
 	 * 'timeout' / 'expire' give us the maximum amount of time
 	 * we wait.
 	 */
-	while (!port->ops->tx_empty(port)) {
+	for (;;) {
+		mutex_lock(&state->port.mutex);
+		if (tty_io_error(tty) || port->ops->tx_empty(port)) {
+			mutex_unlock(&state->port.mutex);
+			break;
+		}
+		mutex_unlock(&state->port.mutex);
+
 		msleep_interruptible(jiffies_to_msecs(char_time));
 		if (signal_pending(current))
 			break;
-- 
2.55.0

[PATCH v3 3/4] serial: fix TIOCMIWAIT race

[Johan Hovold]

The task state must be updated before checking the wakeup condition to
avoid missing a racing modem status update.

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Reported-by: sashiko-bot@kernel.org
Link: https://lore.kernel.org/r/20260904115542.E20D11F00A3D@smtp.kernel.org
Cc: stable@vger.kernel.org
Signed-off-by: Johan Hovold <johan@kernel.org>
---
 drivers/tty/serial/serial_core.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/tty/serial/serial_core.c b/drivers/tty/serial/serial_core.c
index 4213fa3dc988..c2c2c8fe1a4c 100644
--- a/drivers/tty/serial/serial_core.c
+++ b/drivers/tty/serial/serial_core.c
@@ -1230,12 +1230,12 @@ static int uart_wait_modem_status(struct tty_struct *tty, struct uart_state *sta
 
 	add_wait_queue(&port->delta_msr_wait, &wait);
 	for (;;) {
+		set_current_state(TASK_INTERRUPTIBLE);
+
 		uart_port_lock_irq(uport);
 		memcpy(&cnow, &uport->icount, sizeof(struct uart_icount));
 		uart_port_unlock_irq(uport);
 
-		set_current_state(TASK_INTERRUPTIBLE);
-
 		if (((arg & TIOCM_RNG) && (cnow.rng != cprev.rng)) ||
 		    ((arg & TIOCM_DSR) && (cnow.dsr != cprev.dsr)) ||
 		    ((arg & TIOCM_CD)  && (cnow.dcd != cprev.dcd)) ||
-- 
2.55.0

[PATCH v3 4/4] serial: abort TIOCMIWAIT on hangup

[Johan Hovold]

Add the missing hangup check to the TIOCMIWAIT wait loop to avoid going
back to sleep (indefinitely) after being woken on hangup.

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Reported-by: sashiko-bot@kernel.org
Link: https://lore.kernel.org/r/20260904115542.E20D11F00A3D@smtp.kernel.org
Cc: stable@vger.kernel.org
Signed-off-by: Johan Hovold <johan@kernel.org>
---
 drivers/tty/serial/serial_core.c | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/drivers/tty/serial/serial_core.c b/drivers/tty/serial/serial_core.c
index c2c2c8fe1a4c..45d160743ea0 100644
--- a/drivers/tty/serial/serial_core.c
+++ b/drivers/tty/serial/serial_core.c
@@ -1244,6 +1244,11 @@ static int uart_wait_modem_status(struct tty_struct *tty, struct uart_state *sta
 			break;
 		}
 
+		if (tty_io_error(tty)) {
+			ret = -EIO;
+			break;
+		}
+
 		schedule();
 
 		/* see if a signal did it */
-- 
2.55.0

Re: [PATCH v3 1/4] serial: revert guards in uart_wait_modem_status()

[Jiri Slaby]

On 07. 09. 26, 8:44, Johan Hovold wrote:
> Mixing scope-based and regular cleanup is discouraged and
> uart_port_deref() used by uart_wait_modem_status() falls in the latter
> category.
> 
> Revert the premature guard conversion in preparation for fixing a hangup
> race.

It's not premature. Just convert uart_port_ref() to guards too. Gotos suck.

> Fixes: 56609c050051 ("serial: serial_core: use guard()s")
> Cc: stable@vger.kernel.org	# 6.18
> Cc: Jiri Slaby (SUSE) <jirislaby@kernel.org>
> Signed-off-by: Johan Hovold <johan@kernel.org>
> ---
>   drivers/tty/serial/serial_core.c | 13 +++++++------
>   1 file changed, 7 insertions(+), 6 deletions(-)
> 
> diff --git a/drivers/tty/serial/serial_core.c b/drivers/tty/serial/serial_core.c
> index 95774b0f1484..1553bc6cbe7b 100644
> --- a/drivers/tty/serial/serial_core.c
> +++ b/drivers/tty/serial/serial_core.c
> @@ -1213,15 +1213,16 @@ static int uart_wait_modem_status(struct uart_state *state, unsigned long arg)
>   	uport = uart_port_ref(state);
>   	if (!uport)
>   		return -EIO;
> -	scoped_guard(uart_port_lock_irq, uport) {
> -		memcpy(&cprev, &uport->icount, sizeof(struct uart_icount));
> -		uart_enable_ms(uport);
> -	}
> +	uart_port_lock_irq(uport);
> +	memcpy(&cprev, &uport->icount, sizeof(struct uart_icount));
> +	uart_enable_ms(uport);
> +	uart_port_unlock_irq(uport);
>   
>   	add_wait_queue(&port->delta_msr_wait, &wait);
>   	for (;;) {
> -		scoped_guard(uart_port_lock_irq, uport)
> -			memcpy(&cnow, &uport->icount, sizeof(struct uart_icount));
> +		uart_port_lock_irq(uport);
> +		memcpy(&cnow, &uport->icount, sizeof(struct uart_icount));
> +		uart_port_unlock_irq(uport);
>   
>   		set_current_state(TASK_INTERRUPTIBLE);
>   


-- 
js
suse labs

Re: [PATCH v3 3/4] serial: fix TIOCMIWAIT race

[sashiko-bot]

> The task state must be updated before checking the wakeup condition to
> avoid missing a racing modem status update.
> 
> Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
> Reported-by: sashiko-bot@kernel.org
> Link: https://lore.kernel.org/r/20260904115542.E20D11F00A3D@smtp.kernel.org
> Cc: stable@vger.kernel.org
> Signed-off-by: Johan Hovold <johan@kernel.org>

Sashiko has reviewed this patch and found no issues. It looks great!

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260907064418.92953-1-johan@kernel.org?part=3

Re: [PATCH v3 4/4] serial: abort TIOCMIWAIT on hangup

[sashiko-bot]

> Add the missing hangup check to the TIOCMIWAIT wait loop to avoid going
> back to sleep (indefinitely) after being woken on hangup.
> 
> Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
> Reported-by: sashiko-bot@kernel.org
> Link: https://lore.kernel.org/r/20260904115542.E20D11F00A3D@smtp.kernel.org
> Cc: stable@vger.kernel.org
> Signed-off-by: Johan Hovold <johan@kernel.org>

Sashiko has reviewed this patch and found no issues. It looks great!

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260907064418.92953-1-johan@kernel.org?part=4

Re: [PATCH v3 1/4] serial: revert guards in uart_wait_modem_status()

[Jiri Slaby]

On 07. 09. 26, 8:47, Jiri Slaby wrote:
> On 07. 09. 26, 8:44, Johan Hovold wrote:
>> Mixing scope-based and regular cleanup is discouraged and
>> uart_port_deref() used by uart_wait_modem_status() falls in the latter
>> category.
>>
>> Revert the premature guard conversion in preparation for fixing a hangup
>> race.
> 
> It's not premature. Just convert uart_port_ref() to guards too. Gotos suck.

(Or don't use gotos and I will.)

-- 
js
suse labs

Re: [PATCH v3 1/4] serial: revert guards in uart_wait_modem_status()

[sashiko-bot]

> Mixing scope-based and regular cleanup is discouraged and
> uart_port_deref() used by uart_wait_modem_status() falls in the latter
> category.
> 
> Revert the premature guard conversion in preparation for fixing a hangup
> race.
> 
> Fixes: 56609c050051 ("serial: serial_core: use guard()s")
> Cc: stable@vger.kernel.org	# 6.18
> Cc: Jiri Slaby (SUSE) <jirislaby@kernel.org>
> Signed-off-by: Johan Hovold <johan@kernel.org>

Sashiko has reviewed this patch and found no issues. It looks great!

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260907064418.92953-1-johan@kernel.org?part=1

Re: [PATCH v3 2/4] serial: fix ioctl hangup race

[sashiko-bot]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:

Pre-existing issues:
- [High] uart_set_ldisc misses the tty_io_error(tty) check while holding port->mutex, allowing hardware access after the port has been shut down.
--

commit 3a1dbe6dd850d3b9aa0ddbed8997abbaa897c8af
Author: Johan Hovold <johan@kernel.org>

    serial: fix ioctl hangup race
    
    The tty ioctls can race with hangup and end up calling into a tty
    driver for a device that is already gone or powered down.
    
    Add the missing checks to make sure the port has not been hung up before
    accessing the hardware to avoid issues like kernel panic due to
    unclocked accesses.

> diff --git a/drivers/tty/serial/serial_core.c b/drivers/tty/serial/serial_core.c
> index 1553bc6cbe7ba..4213fa3dc988b 100644
> --- a/drivers/tty/serial/serial_core.c
> +++ b/drivers/tty/serial/serial_core.c

[ ... ]

> @@ -1647,7 +1658,7 @@ static void uart_set_termios(struct tty_struct *tty,
>  	guard(mutex)(&state->port.mutex);
>  
>  	uport = uart_port_check(state);
> -	if (!uport)
> +	if (!uport || tty_io_error(tty))
>  		return;
>  
>  	/*

[Severity: High]
This isn't a bug introduced by this patch, but should a similar check be
added to uart_set_ldisc?

uart_set_ldisc() also acquires port->mutex and checks if the port is valid,
but it does not check tty_io_error(tty) before calling uport->ops->set_ldisc:

static void uart_set_ldisc(struct tty_struct *tty)
{
        ...
        guard(mutex)(&state->port.mutex);
        uport = uart_port_check(state);
        if (uport && uport->ops->set_ldisc)
                uport->ops->set_ldisc(uport, &tty->termios);
}

If a privileged user executes TIOCSERCONFIG concurrently with a TIOCSETD ioctl,
uart_do_autoconfig() can acquire the mutex, call uart_shutdown() to shut down
the hardware, and set TTY_IO_ERROR.  If uart_set_ldisc() then acquires the
mutex, could it invoke the uport->ops->set_ldisc hardware callback on a
powered-down device since it doesn't verify tty_io_error(tty)?

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260907064418.92953-1-johan@kernel.org?part=2

Re: [PATCH v3 2/4] serial: fix ioctl hangup race

[Johan Hovold]

On Mon, Sep 07, 2026 at 07:00:34AM +0000, sashiko-bot@kernel.org wrote:
> Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
> 
> Pre-existing issues:
> - [High] uart_set_ldisc misses the tty_io_error(tty) check while holding port->mutex, allowing hardware access after the port has been shut down.

This looks valid so I'll include uart_set_ldisc() in a v4 in a couple of
days.

Johan

Re: [PATCH v3 1/4] serial: revert guards in uart_wait_modem_status()

[Johan Hovold]

On Mon, Sep 07, 2026 at 08:51:26AM +0200, Jiri Slaby wrote:
> On 07. 09. 26, 8:47, Jiri Slaby wrote:
> > On 07. 09. 26, 8:44, Johan Hovold wrote:
> >> Mixing scope-based and regular cleanup is discouraged and
> >> uart_port_deref() used by uart_wait_modem_status() falls in the latter
> >> category.
> >>
> >> Revert the premature guard conversion in preparation for fixing a hangup
> >> race.
> > 
> > It's not premature. Just convert uart_port_ref() to guards too. Gotos suck.
> 
> (Or don't use gotos and I will.)

I guess you meant "don't use guards" here. :)

Gotos don't suck. Obscure driver-specific guard constructs do, though.

So I'm with netdev here, who discourage its use. [1]

I haven't missed your fondness of them, though, and if you want to
continue that path I think it's best if you do it as a follow-on change
(e.g. to facilitate backporting and as it is a larger change to the
driver, unrelated to the fix).

Johan


[1] https://docs.kernel.org/process/maintainer-netdev.html#using-device-managed-and-cleanup-h-constructs