Re: [PATCH 1/1] TTY: don't allow reopen when ldisc is changing

[Kyle McMartin]

On Fri, Nov 26, 2010 at 08:46:18AM +0100, Jiri Slaby wrote:
> >> -	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 &&
> > 
> > Unfortunately, users report this doesn't seem to fix things for them
> > (built against 2.6.36 (plus another patch you wrote iirc.))
> 
> Which patches exactly do you have? You need three of mine in 2.6.36.
> 
> > https://bugzilla.redhat.com/show_bug.cgi?id=630464#c27
> 

Hrm, I'm still seeing it on top of Linus' latest with that patch. :/

Even more bizarrely, I tried to come up with ways this could be failing,
and decided to test a few things...

I set_bit(TTY_DEBUG, &tty->flags) just before returning from
tty_init_dev (which, afaict, should be called for vc/tty$n and ptys?)
and then checked it with a similar WARN_ON in tty_reopen, and found that
I was hitting it fairly regularly.

As far as I can tell, for this to occur, we'd need something to open
/dev/tty1 first, which hits the tty_init_dev, and something else to very
closely follow that, hit the linking of driver->ttys[idx] and so skip
into tty_reopen, and smack into my WARN_ON.

Of course, given the locking, I have no idea how it could possibly be
happening.

I'm poking around to see, I think maybe something might be dropping
locks in the callchain that gives us a window where this might be
possible... I don't see any other way we could end up with tty1 having
TTY_LDISC unset.

(I'm poking in some more debugging, and moving the 'linking in' of the
 device until after tty_ldisc_setup in tty_init_dev, but I'm not
 particularly hopeful.)

 --Kyle
--
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/