Re: tty: AB-BA between tty->legacy_mutex and devpts_mutex

[Jiri Kosina]

On Mon, 28 May 2012, Jiri Kosina wrote:

> ======================================================
> [ INFO: possible circular locking dependency detected ]
> 3.4.0-08219-g238d69d #11 Not tainted
> -------------------------------------------------------
> blogd/265 is trying to acquire lock:
>  (devpts_mutex){+.+.+.}, at: [<ffffffff8134d696>] pty_close+0x166/0x190
> 
> but task is already holding lock:
>  (&tty->legacy_mutex){+.+.+.}, at: [<ffffffff8150f632>] tty_lock+0x22/0x29
> 
> which lock already depends on the new lock.

[ ... snip ... ]

> This deadlock scenario doesn't really seem realistic, as it's between 
> open()/close(). Not being really familiar with tty layer, I am not sure 
> what the proper lock ordering in this case is, I am just reporting for you 
> guys to decide how to get rid of this one.

This seems to have been caused by

commit d739e65bb21d34f0f5d3bf4048410e534fbec148
Author: Alan Cox <alan@xxxxxxxxxxxxxxx>
Date:   Thu May 3 22:22:09 2012 +0100

    pty: Lock the devpts bits privately



as it introduces the legacy_mutex -> devpts_mutex dependency here:



@@ -54,8 +56,11 @@ static void pty_close(struct tty_struct *tty, struct file *filp)
        if (tty->driver->subtype == PTY_TYPE_MASTER) {
                set_bit(TTY_OTHER_CLOSED, &tty->flags);
 #ifdef CONFIG_UNIX98_PTYS
-               if (tty->driver == ptm_driver)
+               if (tty->driver == ptm_driver) {
+                       mutex_lock(&devpts_mutex);
                        devpts_pty_kill(tty->link);
+                       mutex_unlock(&devpts_mutex);
+               }
 #endif
                tty_unlock();
                tty_vhangup(tty->link);

-- 
Jiri Kosina
SUSE Labs
--
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/