Re: BUG: tty_insert_flip_string_fixed_flag, unable to handle kernelNULL pointer dereference at 00000004

From: Alan Cox
Date: Tue Jun 05 2012 - 19:23:39 EST


> [ 5350.556089] [<c027adc4>] pty_write+0x2c/0x4c
> [ 5350.556089] [<c0277a44>] n_tty_write+0x24e/0x2d6
> [ 5350.556089] [<c0144a2b>] ? try_to_wake_up+0x18c/0x18c

Basically the other end hung up just as it went to write.

It's a known race. In the current codebase it's very hard to fix up nicely
as we have to drop tty_lock during a pty close or we deadlock. It's very
hard to hit but your environment sounds rather like my test case for it !

It's one of a series of related bugs (another in n_gsm) that we can't fix
elegantly until the tty locking is fixed - which has been pushed from 3.5
hopefully to 3.6 assuming we can fix a couple of other locking problems
first.

What may reduce it a lot is to change pty_write to do


struct tty_struct *to = tty->link;

if (tty->stopped)
return 0;

if (to == NULL)
return -ENODEV;
to = tty_kref_get(to);
if (tty->link == NULL) {
tty_kref_put(to);
return -ENODEV;
}
if (c > 0) {
...
}
tty_kref_put(to);
return c;
}

Thats a pretty ugly bandaid but I'd be interested to know if it helps, as
that would verify the race you are hitting is the one I think it is.

Alan
--
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/