linux-kernel - Re: [PATCH v2] serial: mxs-auart: add spinlock around changing cts state

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Date: Tue, 19 Mar 2024 14:22:24 -0400
From: Frank Li <Frank.li@....com>
To: Emil Kronborg <emil.kronborg@...tonmail.com>
Cc: jirislaby@...nel.org, gregkh@...uxfoundation.org,
	linux-kernel@...r.kernel.org, linux-serial@...r.kernel.org,
	linux-arm-kernel@...ts.infradead.org, imx@...ts.linux.dev,
	stable@...r.kernel.org
Subject: Re: [PATCH v2] serial: mxs-auart: add spinlock around changing cts
 state

On Tue, Mar 19, 2024 at 06:02:47PM +0000, Emil Kronborg wrote:
> The uart_handle_cts_change() function in serial_core expects the caller
> to hold uport->lock. For example, I have seen the below kernel splat,
> when the Bluetooth driver is loaded on an i.MX28 board.
> 
>     [   85.119255] ------------[ cut here ]------------
>     [   85.124413] WARNING: CPU: 0 PID: 27 at /drivers/tty/serial/serial_core.c:3453 uart_handle_cts_change+0xb4/0xec
>     [   85.134694] Modules linked in: hci_uart bluetooth ecdh_generic ecc wlcore_sdio configfs
>     [   85.143314] CPU: 0 PID: 27 Comm: kworker/u3:0 Not tainted 6.6.3-00021-gd62a2f068f92 #1
>     [   85.151396] Hardware name: Freescale MXS (Device Tree)
>     [   85.156679] Workqueue: hci0 hci_power_on [bluetooth]
>     (...)
>     [   85.191765]  uart_handle_cts_change from mxs_auart_irq_handle+0x380/0x3f4
>     [   85.198787]  mxs_auart_irq_handle from __handle_irq_event_percpu+0x88/0x210
>     (...)
> 
> Cc: stable@...r.kernel.org # v6.1+
> Fixes: 4d90bb147ef6 ("serial: core: Document and assert lock requirements for irq helpers")
> Signed-off-by: Emil Kronborg <emil.kronborg@...tonmail.com>

Reviewed-by: Frank Li <Frank.Li@....com>

> ---
> Changes in v2:
> - Add more of the stack trace to show why the lock is necessary. Note
>   that i removed some unrelated noise for unwinding, showing and dumping
>   the stack trace.
> - Add Fixes tag. Note that this commit do not fix 4d90bb147ef6 ("serial:
>   core: Document and assert lock requirements for irq helpers") as such,
>   but it was the closest commit I could find that makes sense.
> - Cc stable. Commit b0af4bcb4946 ("serial: core: Provide port lock
>   wrappers") is a prerequisite. Therefore, v6.1+.
> 
>  drivers/tty/serial/mxs-auart.c | 8 ++++++--
>  1 file changed, 6 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/tty/serial/mxs-auart.c b/drivers/tty/serial/mxs-auart.c
> index 4749331fe618..1e8853eae504 100644
> --- a/drivers/tty/serial/mxs-auart.c
> +++ b/drivers/tty/serial/mxs-auart.c
> @@ -1086,11 +1086,13 @@ static void mxs_auart_set_ldisc(struct uart_port *port,
>  
>  static irqreturn_t mxs_auart_irq_handle(int irq, void *context)
>  {
> -	u32 istat;
> +	u32 istat, stat;
>  	struct mxs_auart_port *s = context;
>  	u32 mctrl_temp = s->mctrl_prev;
> -	u32 stat = mxs_read(s, REG_STAT);
>  
> +	uart_port_lock(&s->port);
> +
> +	stat = mxs_read(s, REG_STAT);
>  	istat = mxs_read(s, REG_INTR);
>  
>  	/* ack irq */
> @@ -1126,6 +1128,8 @@ static irqreturn_t mxs_auart_irq_handle(int irq, void *context)
>  		istat &= ~AUART_INTR_TXIS;
>  	}
>  
> +	uart_port_unlock(&s->port);
> +
>  	return IRQ_HANDLED;
>  }
>  
> -- 
> 2.44.0
> 
>