Skip to contentSkip to conversation

tty: serial: core: fix NULL pointer deref in uart_resume_port()

PATCH In review Weiming Shi 2 participants 3 messages Jun 8 – Jun 13 TTY layer and serial drivers

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 tty: serial: core: fix NULL pointer deref in uart_resume_port()

uart_resume_port() looks up the tty device child with device_find_child() and passes the result straight to device_may_wakeup(). device_find_child() returns NULL when the port has no matching tty device child, and device_may_wakeup() dereferences dev->power.can_wakeup, so a NULL tty_dev faults. uart_suspend_port() already guards the same call with "tty_dev && device_may_wakeup(tty_dev)"; the resume path does not.

 Oops: general protection fault, probably for non-canonical address
 KASAN: null-ptr-deref in range [0x148-0x14f]
 RIP: 0010:uart_resume_port (pm_wakeup.h:84 serial_core.c:2477)
  serial_pnp_resume (8250/8250_pnp.c:522)
  pnp_bus_resume (drivers/pnp/driver.c:234)

Mirror the NULL guard from uart_suspend_port(). put_device(tty_dev) already tolerates a NULL argument, so only the device_may_wakeup() call needs the check; the non-NULL path is unchanged.

Fixes b3b708fa2780 ("wake up from a serial port")
Assisted-by Claude:claude-opus-4-8
---
 drivers/tty/serial/serial_core.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/tty/serial/serial_core.c +1 -1
index 89cebdd278410..7dd3ec3dea438 100644
@@ -2373,7 +2373,7 @@ int uart_resume_port(struct uart_driver *drv, struct uart_port *uport)
 	guard(mutex)(&port->mutex);
 
 	tty_dev = device_find_child(&uport->port_dev->dev, &match, serial_match_port);
-	if (!uport->suspended && device_may_wakeup(tty_dev)) {
+	if (!uport->suspended && tty_dev && device_may_wakeup(tty_dev)) {
 		if (irqd_is_wakeup_set(irq_get_irq_data((uport->irq))))
 			disable_irq_wake(uport->irq);
 		put_device(tty_dev);
Signature · 1 line
2.43.0
#1 lore raw
Quoted 3 lines · Weiming Shi

uart_resume_port() looks up the tty device child with device_find_child() and passes the result straight to device_may_wakeup(). device_find_child() returns NULL when the port has no matching tty device child,

How can that happen in a real system? Have you triggered this before, if so, what hardware does it?

Quoted 10 lines

and
device_may_wakeup() dereferences dev->power.can_wakeup, so a NULL tty_dev faults. uart_suspend_port() already guards the same call with "tty_dev && device_may_wakeup(tty_dev)"; the resume path does not.

 Oops: general protection fault, probably for non-canonical address
 KASAN: null-ptr-deref in range [0x148-0x14f]
 RIP: 0010:uart_resume_port (pm_wakeup.h:84 serial_core.c:2477)
  serial_pnp_resume (8250/8250_pnp.c:522)
  pnp_bus_resume (drivers/pnp/driver.c:234)

Is this a real oops, or a made up one?

Quoted 6 lines

Mirror the NULL guard from uart_suspend_port(). put_device(tty_dev) already tolerates a NULL argument, so only the device_may_wakeup() call needs the check; the non-NULL path is unchanged.

Fixes b3b708fa2780 ("wake up from a serial port")

Where was this reported?

Why isn't this cc: stable? And why hasn't anyone tripped over it in the past 19 years?

thanks,

greg k-h

#2 lore raw
Quoted 28 lines · Greg Kroah-Hartman
Quoted 3 lines · Weiming Shi

uart_resume_port() looks up the tty device child with device_find_child() and passes the result straight to device_may_wakeup(). device_find_child() returns NULL when the port has no matching tty device child,

How can that happen in a real system? Have you triggered this before, if so, what hardware does it?

Quoted 10 lines

and
device_may_wakeup() dereferences dev->power.can_wakeup, so a NULL tty_dev faults. uart_suspend_port() already guards the same call with "tty_dev && device_may_wakeup(tty_dev)"; the resume path does not.

 Oops: general protection fault, probably for non-canonical address
 KASAN: null-ptr-deref in range [0x148-0x14f]
 RIP: 0010:uart_resume_port (pm_wakeup.h:84 serial_core.c:2477)
  serial_pnp_resume (8250/8250_pnp.c:522)
  pnp_bus_resume (drivers/pnp/driver.c:234)

Is this a real oops, or a made up one?

Quoted 6 lines

Mirror the NULL guard from uart_suspend_port(). put_device(tty_dev) already tolerates a NULL argument, so only the device_may_wakeup() call needs the check; the non-NULL path is unchanged.

Fixes b3b708fa2780 ("wake up from a serial port")
Quoted 8 lines

Where was this reported?

Why isn't this cc: stable? And why hasn't anyone tripped over it in the past 19 years?

thanks,

greg k-h

Hi greg,

This is a false positive, please drop it.
Sorry for wasting your time. I'll check reachability before sending anything next time.

Best,
Weiming Shi