[PATCH] tty: serial: 21285: fix lockup on open

[PATCH] tty: serial: 21285: fix lockup on open

[Russell King]

Commit 293f89959483 ("tty: serial: 21285: stop using the unused[]
variable from struct uart_port") introduced a bug which stops the
transmit interrupt being disabled when there are no characters to
transmit - disabling the transmit interrupt at the interrupt controller
is the only way to stop an interrupt storm. If this interrupt is not
disabled when there are no transmit characters, we end up with an
interrupt storm which prevents the machine making forward progress.

Fixes: 293f89959483 ("tty: serial: 21285: stop using the unused[] variable from struct uart_port")
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>
---
 drivers/tty/serial/21285.c | 12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)

diff --git a/drivers/tty/serial/21285.c b/drivers/tty/serial/21285.c
index 718e010fcb04..09baef4ccc39 100644
--- a/drivers/tty/serial/21285.c
+++ b/drivers/tty/serial/21285.c
@@ -50,25 +50,25 @@ static const char serial21285_name[] = "Footbridge UART";
 
 static bool is_enabled(struct uart_port *port, int bit)
 {
-	unsigned long private_data = (unsigned long)port->private_data;
+	unsigned long *private_data = (unsigned long *)&port->private_data;
 
-	if (test_bit(bit, &private_data))
+	if (test_bit(bit, private_data))
 		return true;
 	return false;
 }
 
 static void enable(struct uart_port *port, int bit)
 {
-	unsigned long private_data = (unsigned long)port->private_data;
+	unsigned long *private_data = (unsigned long *)&port->private_data;
 
-	set_bit(bit, &private_data);
+	set_bit(bit, private_data);
 }
 
 static void disable(struct uart_port *port, int bit)
 {
-	unsigned long private_data = (unsigned long)port->private_data;
+	unsigned long *private_data = (unsigned long *)&port->private_data;
 
-	clear_bit(bit, &private_data);
+	clear_bit(bit, private_data);
 }
 
 #define is_tx_enabled(port)	is_enabled(port, tx_enabled_bit)
-- 
2.20.1

Re: [PATCH] tty: serial: 21285: fix lockup on open

[Greg Kroah-Hartman]

On Sun, Oct 18, 2020 at 09:42:04AM +0100, Russell King wrote:
> Commit 293f89959483 ("tty: serial: 21285: stop using the unused[]
> variable from struct uart_port") introduced a bug which stops the
> transmit interrupt being disabled when there are no characters to
> transmit - disabling the transmit interrupt at the interrupt controller
> is the only way to stop an interrupt storm. If this interrupt is not
> disabled when there are no transmit characters, we end up with an
> interrupt storm which prevents the machine making forward progress.
> 
> Fixes: 293f89959483 ("tty: serial: 21285: stop using the unused[] variable from struct uart_port")
> Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>
> ---
>  drivers/tty/serial/21285.c | 12 ++++++------
>  1 file changed, 6 insertions(+), 6 deletions(-)
> 
> diff --git a/drivers/tty/serial/21285.c b/drivers/tty/serial/21285.c
> index 718e010fcb04..09baef4ccc39 100644
> --- a/drivers/tty/serial/21285.c
> +++ b/drivers/tty/serial/21285.c
> @@ -50,25 +50,25 @@ static const char serial21285_name[] = "Footbridge UART";
>  
>  static bool is_enabled(struct uart_port *port, int bit)
>  {
> -	unsigned long private_data = (unsigned long)port->private_data;
> +	unsigned long *private_data = (unsigned long *)&port->private_data;
>  
> -	if (test_bit(bit, &private_data))
> +	if (test_bit(bit, private_data))
>  		return true;
>  	return false;
>  }
>  
>  static void enable(struct uart_port *port, int bit)
>  {
> -	unsigned long private_data = (unsigned long)port->private_data;
> +	unsigned long *private_data = (unsigned long *)&port->private_data;
>  
> -	set_bit(bit, &private_data);
> +	set_bit(bit, private_data);
>  }
>  
>  static void disable(struct uart_port *port, int bit)
>  {
> -	unsigned long private_data = (unsigned long)port->private_data;
> +	unsigned long *private_data = (unsigned long *)&port->private_data;
>  
> -	clear_bit(bit, &private_data);
> +	clear_bit(bit, private_data);
>  }
>  
>  #define is_tx_enabled(port)	is_enabled(port, tx_enabled_bit)
> -- 
> 2.20.1
> 

Sorry about this, my fault.  I'll merge this after 5.10-rc1 is out,
thanks for the fix.

greg k-h