Re: [PATCH v5] bluetooth: Fix WARNING in tty_set_termios()
From: Johan Hovold
Date: Sat Feb 09 2019 - 04:42:50 EST
On Fri, Feb 08, 2019 at 04:06:09PM -0700, Shuah Khan wrote:
> tty_set_termios() has the following WARN_ON which can be triggered with a
> syscall to invoke TIOCSETD __NR_ioctl.
> WARN_ON(tty->driver->type == TTY_DRIVER_TYPE_PTY &&
> tty->driver->subtype == PTY_TYPE_MASTER);
> Reference: https://syzkaller.appspot.com/bug?id=2410d22f1d8e5984217329dd0884b01d99e3e48d
> "The problem started with commit 7721383f4199 ("Bluetooth: hci_uart: Support
> operational speed during setup") which introduced a new way for how
> tty_set_termios() could end up being called for a master pty."
> Fix hci_uart_tty_open() to check if tty supports set_termios in addition
> to write and error out, if it doesn't.
> Reported-by: syzbot+a950165cbb86bdd023a4@xxxxxxxxxxxxxxxxxxxxxxxxx
> Cc: Johan Hovold <johan@xxxxxxxxxx>
> Cc: Marcel Holtmann <marcel@xxxxxxxxxxxx>
> Cc: Al Viro <viro@xxxxxxxxxxxxxxxxxx>
> Signed-off-by: Shuah Khan <shuah@xxxxxxxxxx>
> drivers/bluetooth/hci_ldisc.c | 6 +++---
> 1 file changed, 3 insertions(+), 3 deletions(-)
> diff --git a/drivers/bluetooth/hci_ldisc.c b/drivers/bluetooth/hci_ldisc.c
> index fbf7b4df23ab..087ec2268744 100644
> --- a/drivers/bluetooth/hci_ldisc.c
> +++ b/drivers/bluetooth/hci_ldisc.c
> @@ -474,10 +474,10 @@ static int hci_uart_tty_open(struct tty_struct *tty)
> BT_DBG("tty %p", tty);
> - /* Error if the tty has no write op instead of leaving an exploitable
> - * hole
> + /* Error if the tty has no write/set_termios ops instead of leaving
> + * an exploitable hole.
Why do you think that a tty driver not implementing set_termios() is
Note that tty_set_termios() handles a missing set_termios() callback
just fine without dereferencing a NULL-pointer.
> - if (tty->ops->write == NULL)
> + if (!tty->ops->write || !tty->ops->set_termios)
> return -EOPNOTSUPP;
I know Marcel asked you do implement things like this, but what you're
really doing is to try to uphold the invariant that tty_set_termios()
should never be called for a pty master. The WARN_ON in that function
was put in place to make sure line-disciplines use the correct ioctl
helpers (and the commit I pointed you to earlier in the paragraph you
quote in the commit message broke that invariant).
I'm afraid this is totally obscured by the above change and misleading