On 2/1/19 3:00 AM, Marcel Holtmann wrote:
Hi Shuah,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 the problem by preventing setting the HCI line discipline for PTYs from hci_uart_setup() and hci_uart_set_flow_control(). The reproducer is used to reproduce the problem and verify the fix. Reported-by: [email protected] Signed-off-by: Shuah Khan <[email protected]> --- drivers/bluetooth/hci_ldisc.c | 11 +++++++++++ 1 file changed, 11 insertions(+) diff --git a/drivers/bluetooth/hci_ldisc.c b/drivers/bluetooth/hci_ldisc.c index fbf7b4df23ab..ce84ca91ca70 100644 --- a/drivers/bluetooth/hci_ldisc.c +++ b/drivers/bluetooth/hci_ldisc.c @@ -314,6 +314,11 @@ void hci_uart_set_flow_control(struct hci_uart *hu, bool enable) return; } + /* don't set HCI line discipline on PTYs */ + if (tty->driver->type == TTY_DRIVER_TYPE_PTY && + tty->driver->subtype == PTY_TYPE_MASTER) + return; +just do it once in hci_uart_tty_open. Actually we already check for ops->write and so lets just ensure all ops we ever call are present. If they are not, then bail out. I wouldn’t even bother with the TTY type. Wouldn’t that be a lot simpler?
As you said, hci_uart_tty_open()is a good place to check it. I think checking tty type is necessary. I couldn't find any missing ops I could base the check on to prevent the ldisc set. I have the patch that does the tty check in hci_uart_tty_open() tested and ready for sending. thanks, -- Shuah
