Re: mmotm 2010-11-23 - WARNING: at drivers/tty/tty_io.c:1331

[Jiri Slaby <jslaby@suse.cz>]

On 11/25/2010 05:44 PM, Jiri Slaby wrote:
> On 11/25/2010 04:14 PM, Kyle McMartin wrote:
>> On Tue, Nov 23, 2010 at 11:55:39PM -0500, Valdis.Kletnieks@vt.edu wrote:
>>> On Tue, 23 Nov 2010 16:13:06 PST, akpm@linux-foundation.org said:
>>>> The mm-of-the-moment snapshot 2010-11-23-16-12 has been uploaded to
>>>>
>>>>    http://userweb.kernel.org/~akpm/mmotm/
>>>
>>> Seen during boot:
>>>
>>> [   23.015448] Modules linked in:
>>> [   23.015453] Pid: 1207, comm: plymouthd Not tainted 2.6.37-rc3-mmotm1123 #3
>>> [   23.015455] Call Trace:
>>
>> I've been trying to figure this one out for a while, without much luck.
>> (Users are seeing it in 2.6.36 as well.)
>>
>> I *think* (I added a rawhide debugging patch to print the tty->name)
>> that plymouth is always opening tty7 to cause this. My guess is the BKL
>> removal has exposed some kind of race, but it's not obvious to me (and
>> there's many other bugs to sort through too. :(
>>
>> CC-ing Jiri since he seems to be the poor guy who's been poking this
>> recently (there's a good few threads about this (though the others look
>> like an ldisc attach race...)) I wouldn't think that's the case here
>> since N_TTY is the default...
> 
> Ok, tty_reopen is called without TTY_LDISC set. For further
> considerations, note tty_lock is held in tty_open. TTY_LDISC is cleared in:
> 
> 1) __tty_hangup from tty_ldisc_hangup to tty_ldisc_enable. During this
> section tty_lock is held.
> 
> 2) tty_release via tty_ldisc_release till the end of tty existence. If
> tty->count <= 1, tty_lock is taken, TTY_CLOSING bit set and then
> tty_ldisc_release called. tty_reopen checks TTY_CLOSING before checking
> TTY_LDISC.
> 
> 3) tty_set_ldisc from tty_ldisc_halt to tty_ldisc_enable. We take
> tty_lock, set TTY_LDISC_CHANGING, put tty_lock, do some other work, take
> tty_lock, call tty_ldisc_enable, put tty_lock.

Oh, "do some other work" includes tty_ldisc_halt where TTY_LDISC is
cleared and tty_lock is _not_ held.

> So the only option I see is 3) and we should do:
> --- a/drivers/tty/tty_io.c
> +++ b/drivers/tty/tty_io.c
> @@ -1310,7 +1310,8 @@ static int tty_reopen(struct tty_struct *tty)
>  {
>         struct tty_driver *driver = tty->driver;
> 
> -       if (test_bit(TTY_CLOSING, &tty->flags))
> +       if (test_bit(TTY_CLOSING, &tty->flags) ||
> +                       test_bit(TTY_LDISC_CHANGING, &tty->flags))
>                 return -EIO;
> 
>         if (driver->type == TTY_DRIVER_TYPE_PTY &&
> 
> Alan, Greg?
> 
> thanks,
-- 
js
suse labs