Re: [syzbot] [kernel?] KCSAN: data-race in __fput / __tty_hangup (4)

[Al Viro]

On Sat, Apr 22, 2023 at 12:12:02AM +0900, Tetsuo Handa wrote:
> On 2023/04/21 17:21, Dmitry Vyukov wrote:
> > If I am reading this correctly, this race can lead to NULL derefs
> > among other things.
> > hung_up_tty_fops does not have splice_read, while other fops have.
> > 
> > So the following code in splice can execute NULL callback:
> > 
> > if (unlikely(!in->f_op->splice_read))
> >     return warn_unsupported(in, "read");
> > return in->f_op->splice_read(in, ppos, pipe, len, flags);
> > 
> 
> __fput(file) is called when the last reference to file is released.
> Since __tty_hangup() traverses tty->tty_files under tty->files_lock,
> tty_add_file() needs to hold a ref before adding to tty->tty_files
> in order to defer concurrent __fput() by other threads?
> 
> diff --git a/drivers/tty/tty_io.c b/drivers/tty/tty_io.c
> index 36fb945fdad4..2838703d48cf 100644
> --- a/drivers/tty/tty_io.c
> +++ b/drivers/tty/tty_io.c
> @@ -197,7 +197,7 @@ void tty_add_file(struct tty_struct *tty, struct file *file)
>  	struct tty_file_private *priv = file->private_data;
>  
>  	priv->tty = tty;
> -	priv->file = file;
> +	priv->file = get_file(file);

This is broken.  Simple open() + close() on a tty will leak the damn thing -
that extra reference will stick around, possibly all the way until reboot.

As for the original report - add a (failing) ->splice_read() in hung_ut_tty_fops
to deal with the original problem.