Re: [Fwd: [PATCH v2 0/4] TTY: port hangup and close fixes]

[Jiri Slaby <jslaby@suse.cz>]

On 03/05/2013 11:02 PM, Peter Hurley wrote:
> On Tue, 2013-03-05 at 22:56 +0100, Jiri Slaby wrote:
>> On 03/05/2013 06:06 PM, Peter Hurley wrote:
>>>>> @@ -225,15 +232,13 @@ void tty_port_hangup(struct tty_port *port)
>>>>         spin_lock_irqsave(&port->lock, flags);
>>>>         port->count = 0;
>>>>         port->flags &= ~ASYNC_NORMAL_ACTIVE;
>>>> -       if (port->tty) {
>>>> +       if (port->tty)
>>>>                 set_bit(TTY_IO_ERROR, &port->tty->flags);
>>>> -               tty_kref_put(port->tty);
>>>> -       }
>>>> -       port->tty = NULL;
>>>>         spin_unlock_irqrestore(&port->lock, flags);
>>>>> +       tty_port_shutdown(port, port->tty);
>>>>
>>>> What prevents port->tty to be NULL here already?
>>>
>>> Nothing. That's why it's tested in tty_port_shutdown() above.
>>
>> I know :).
> 
> Sorry :)
> 
>> But the question is rather don't we want to pass the real
>> refcounted port->tty (take a snapshot inside the lock) instead?
> 
> I think that's why he moved the kref release to after the shutdown (via
> tty_port_set_tty()) -- but I'm tired and maybe I'm missing something
> here?

port->tty can be changed right after the unlock. So I'm thinking about
something like this:

if (port->tty)
   set_bit(TTY_IO_ERROR, &port->tty->flags);
tty = port->tty; <=== take a snapshot
spin_unlock_irqrestore(&port->lock, flags);
tty_port_shutdown(port, tty); <=== use the snapshot
set_tty_port(port, NULL); <=== put kref on that tty

-- 
js
suse labs