Re: BUG: NULL pointer deref in tty port / uart

[Jiri Olsa]

On Wed, May 18, 2011 at 12:44:20AM +0100, Alan Cox wrote:
> > 	echo 1 > /dev/ttyS4 # which blocks
> > 
> > And on another console:
> > 
> > 	cat /dev/ttyS4 # which blocks
> > 
> > Then Ctrl + C the echo in the first console. This produces the
> > following trace:
> 
> First cat is in tty_port_block_til_ready, second cat joins it there. ^C
> causes one to close, which wakes the second which goes around the loop
> again, tries to raise the carrier and explodes, it seems because
> someone trashed memory it is using.
> 
> Not quite sure why at this point
> 
> On the first exit of the open path port->count is 1 which is as we want
> it. Close takes it down to zero which triggers the port shutdown path
> which is as we want. We clean up port->tty and shut down the port.
> Seeing the second pending open we wake it which is when it goes kaboom
> 
> Nothing obvious strikes me from reading the code.

hi,

have the same issue.. looks like we should not NULL the port->tty
if there's blocked open, but not sure what's exactly the logic
behind "port's block_open and count" ..

attached patch fixes it for me

wbr,
jirka

---
 drivers/tty/serial/serial_core.c |    4 +++-
 1 files changed, 3 insertions(+), 1 deletions(-)

diff --git a/drivers/tty/serial/serial_core.c b/drivers/tty/serial/serial_core.c
index 733fe8e..86a40cb 100644
--- a/drivers/tty/serial/serial_core.c
+++ b/drivers/tty/serial/serial_core.c
@@ -1346,7 +1346,9 @@ static void uart_close(struct tty_struct *tty, struct file *filp)
 
 	tty_ldisc_flush(tty);
 
-	tty_port_tty_set(port, NULL);
+	if (!tty_port_users(port))
+		tty_port_tty_set(port, NULL);
+
 	spin_lock_irqsave(&port->lock, flags);
 	tty->closing = 0;
 
-- 
1.7.1
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/