Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Tue, 25 Apr 2023 17:03:44 +0100
From:   Al Viro <viro@zeniv.linux.org.uk>
To:     Tetsuo Handa <penguin-kernel@i-love.sakura.ne.jp>
Cc:     Jiri Slaby <jirislaby@kernel.org>,
        Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
        Dmitry Vyukov <dvyukov@google.com>,
        syzbot <syzbot+b7c3ba8cdc2f6cf83c21@syzkaller.appspotmail.com>,
        linux-kernel@vger.kernel.org, syzkaller-bugs@googlegroups.com
Subject: Re: [syzbot] [kernel?] KCSAN: data-race in __fput / __tty_hangup (4)
Message-ID: <20230425160344.GS3390869@ZenIV>
References: <00000000000013aaac05f9d44e7a@google.com>
 <CACT4Y+bj7jCOjGV5jCUPzN5zjgdHxRn9vkwQEBuPWVzMbMCnXw@mail.gmail.com>
 <6ff6fdea-c955-f9dd-289e-b0d613a28280@I-love.SAKURA.ne.jp>
 <20230423233433.GF3390869@ZenIV>
 <e1fe6a44-3021-62ad-690a-69146e39e1ac@I-love.SAKURA.ne.jp>
 <20230424004431.GG3390869@ZenIV>
 <8e21256a-736e-4c2d-1ff4-723775bcac46@I-love.SAKURA.ne.jp>
 <2fca7932-5030-32c3-dd61-48dd78e58e11@I-love.SAKURA.ne.jp>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <2fca7932-5030-32c3-dd61-48dd78e58e11@I-love.SAKURA.ne.jp>
Sender: Al Viro <viro@ftp.linux.org.uk>
Precedence: bulk

On Tue, Apr 25, 2023 at 11:47:20PM +0900, Tetsuo Handa wrote:
> @@ -743,7 +733,8 @@ void tty_vhangup_session(struct tty_struct *tty)
>   */
>  int tty_hung_up_p(struct file *filp)
>  {
> -	return (filp && filp->f_op == &hung_up_tty_fops);
> +	/* Accept race with __tty_hangup(). */
> +	return filp && data_race(((struct tty_file_private *) filp->private_data)->hung);
>  }
>  EXPORT_SYMBOL(tty_hung_up_p);

Umm...  Are you sure we never call that for non-TTY files?  Seeing that
it's exported and all such...  For internal uses (tty_read(), etc.) the
check for file being NULL is pointless; for general-purpose primitive
we probably want to check ->f_op as well...

> @@ -2159,11 +2154,6 @@ static int tty_open(struct inode *inode, struct file *filp)
>  			return retval;
>  
>  		schedule();
> -		/*
> -		 * Need to reset f_op in case a hangup happened.
> -		 */
> -		if (tty_hung_up_p(filp))
> -			filp->f_op = &tty_fops;

That needs a comment in commit message - hangup indication doesn't
need to be reset here, since tty_release() has freed the entire thing
that used to hang off the ->private_data.

> I don't know what changes are required for tty_open() and tty_show_fdinfo().
> I assume that making no change for tty_show_fdinfo() is harmless.
> But how does "hung_up_tty_fops does not call tty_open()" affects here?

It doesn't - file->f_op->open() is only called once, and only on fresh
struct file.