[patch] Staging: generic_serial: fix double locking bug

[patch] Staging: generic_serial: fix double locking bug

[Dan Carpenter]

spin_lock_irqsave() is not nestable.  The second time that it gets
called it overwrites the "flags" variable and so IRQs can't be restored
properly.

Signed-off-by: Dan Carpenter <error27@gmail.com>
---
Oh.  Crap.  I don't want to adopt these drivers.  But I already wrote
the patch before I read the TODO file, so I'm still going to submit it.

diff --git a/drivers/staging/generic_serial/generic_serial.c b/drivers/staging/generic_serial/generic_serial.c
index 5954ee1..466988d 100644
--- a/drivers/staging/generic_serial/generic_serial.c
+++ b/drivers/staging/generic_serial/generic_serial.c
@@ -566,9 +566,9 @@ void gs_close(struct tty_struct * tty, struct file * filp)
 	 * line status register.
 	 */
 
-	spin_lock_irqsave(&port->driver_lock, flags);
+	spin_lock(&port->driver_lock);
 	port->rd->disable_rx_interrupts (port);
-	spin_unlock_irqrestore(&port->driver_lock, flags);
+	spin_unlock(&port->driver_lock);
 	spin_unlock_irqrestore(&port->port.lock, flags);
 
 	/* close has no way of returning "EINTR", so discard return value */

Re: [patch] Staging: generic_serial: fix double locking bug

[Greg KH]

On Mon, Feb 28, 2011 at 10:28:31PM +0300, Dan Carpenter wrote:
> spin_lock_irqsave() is not nestable.  The second time that it gets
> called it overwrites the "flags" variable and so IRQs can't be restored
> properly.
> 
> Signed-off-by: Dan Carpenter <error27@gmail.com>
> ---
> Oh.  Crap.  I don't want to adopt these drivers.  But I already wrote
> the patch before I read the TODO file, so I'm still going to submit it.

Heh, don't worry, you don't "own" them, I'll just take this fix for now
and still leave them in the "abandoned" category :)

thanks,

greg k-h