Re: [PATCH] serial: imx: cancel RS485 trigger hrtimers in shutdown and remove

[sashiko-bot@kernel.org]

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

Pre-existing issues:
- [High] Missing cancellation of `sport->timer` and modem status GPIO IRQs in `imx_uart_remove()` leads to a use-after-free and system crash (bus fault) for console ports.
--

--- Patch [1]: [PATCH] serial: imx: cancel RS485 trigger hrtimers in shutdown and remove ---
commit 00c3fef931cf8afdbbbb1c6e8dc49797a632d082
Author: Fan Wu <fanwu01@zju.edu.cn>

serial: imx: cancel RS485 trigger hrtimers in shutdown and remove

This commit cancels the RS485 trigger hrtimers synchronously in
imx_uart_shutdown() and imx_uart_remove() to prevent use-after-free
after the devm-allocated port is freed.

> diff --git a/drivers/tty/serial/imx.c b/drivers/tty/serial/imx.c
> index 251a50c8aa387..86c99f73c50a3 100644
> --- a/drivers/tty/serial/imx.c
> +++ b/drivers/tty/serial/imx.c

[ ... ]

> @@ -2649,6 +2653,10 @@ static void imx_uart_remove(struct platform_device *pdev)
>  	struct imx_port *sport = platform_get_drvdata(pdev);
>  
>  	uart_remove_one_port(&imx_uart_uart_driver, &sport->port);
> +
> +	/* Serial core can reach remove() without calling the driver shutdown. */
> +	hrtimer_cancel(&sport->trigger_start_tx);
> +	hrtimer_cancel(&sport->trigger_stop_tx);

[Severity: High]
This is a pre-existing issue, but since we are cancelling timers in
imx_uart_remove(), does this function also need to cancel sport->timer and
disable the modem status GPIO IRQs?

If the system boots with hardware flow control enabled on the console,
imx_uart_console_setup() sets termios, which calls imx_uart_enable_ms() and
arms sport->timer.

If the driver is unbound via sysfs or the module is unloaded,
imx_uart_remove() unregisters the console, disabling the UART clocks.
However, it does not cancel sport->timer or disable the MS GPIO interrupts.

When devres frees the imx_port structure, the timer can still fire:

drivers/tty/serial/imx.c:imx_uart_timeout() {
        struct imx_port *sport = timer_container_of(sport, t, timer);
        unsigned long flags;

        if (sport->port.state) {
                uart_port_lock_irqsave(&sport->port, &flags);
                imx_uart_mctrl_check(sport);
                uart_port_unlock_irqrestore(&sport->port, flags);

                mod_timer(&sport->timer, jiffies + MCTRL_TIMEOUT);
        }
}

Could this lead to accessing the freed sport struct and reading disabled
UART registers, resulting in a use-after-free and system crash?

>  }
>  
>  static void imx_uart_restore_context(struct imx_port *sport)

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260819021916.442827-1-fanwu01@zju.edu.cn?part=1