From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1754026AbZHYAK6 (ORCPT ); Mon, 24 Aug 2009 20:10:58 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1753997AbZHYAKz (ORCPT ); Mon, 24 Aug 2009 20:10:55 -0400 Received: from smtp1.linux-foundation.org ([140.211.169.13]:52690 "EHLO smtp1.linux-foundation.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753671AbZHYAKz (ORCPT ); Mon, 24 Aug 2009 20:10:55 -0400 Date: Mon, 24 Aug 2009 17:09:08 -0700 (PDT) From: Linus Torvalds X-X-Sender: torvalds@localhost.localdomain To: "Eric W. Biederman" cc: linux-kernel@vger.kernel.org, x86@kernel.org, Thomas Gleixner , Ingo Molnar , "H. Peter Anvin" , Alan Cox , Greg Kroah-Hartman Subject: Re: v2.6.31-rc6: BUG: unable to handle kernel NULL pointer dereference at 0000000000000008 In-Reply-To: Message-ID: References: User-Agent: Alpine 2.01 (LFD 1184 2008-12-16) MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Mon, 24 Aug 2009, Linus Torvalds wrote: > > But I wanted to let people know that the patch is clearly not the "last > word" on this. It's a useful thing to try, but we need something better. This may be better (this is a replacement for the previous patch). Instead of using 'cancel_delayed_work_sync()', it makes tty_ldisc_hangup() do a 'flush_scheduled_work()' afterwards, like the other callers already do. And like 'tty_ldisc_release()' already does, it does this all before even getting the ldisc_mutex, avoiding the deadlock. I'm not 100% happy with this patch either, but my remaining unhappiness is more with the tty locking in general that causes this all. I suspect this patch in itself is not any worse than the other hacks we have. Oh, and in case you didn't guess - this is _STILL_ totally untested. It compiles for me, but that's all I'm going to guarantee. I'm just looking at the code (and getting pretty fed up with it ;) And as already mentioned: I doubt the deadlock on tty->ldisc_mutex is anything that would be hit in practice. And even if it can be triggered, the previous patch I sent out is still interesting in a "does it make the problem go away" sense. Because if it doesn't (with or without a new deadlock), then I'm looking at all the wrong places. Linus --- drivers/char/tty_ldisc.c | 10 +++++++--- 1 files changed, 7 insertions(+), 3 deletions(-) diff --git a/drivers/char/tty_ldisc.c b/drivers/char/tty_ldisc.c index 1733d34..f893d18 100644 --- a/drivers/char/tty_ldisc.c +++ b/drivers/char/tty_ldisc.c @@ -508,8 +508,9 @@ static void tty_ldisc_restore(struct tty_struct *tty, struct tty_ldisc *old) * be obtained while the delayed work queue halt ensures that no more * data is fed to the ldisc. * - * In order to wait for any existing references to complete see - * tty_ldisc_wait_idle. + * You need to do a 'flush_scheduled_work()' (outside the ldisc_mutex + * in order to make sure any currently executing ldisc work is also + * flushed. */ static int tty_ldisc_halt(struct tty_struct *tty) @@ -753,11 +754,14 @@ void tty_ldisc_hangup(struct tty_struct *tty) * N_TTY. */ if (tty->driver->flags & TTY_DRIVER_RESET_TERMIOS) { + /* Make sure the old ldisc is quiescent */ + tty_ldisc_halt(tty); + flush_scheduled_work(); + /* Avoid racing set_ldisc or tty_ldisc_release */ mutex_lock(&tty->ldisc_mutex); if (tty->ldisc) { /* Not yet closed */ /* Switch back to N_TTY */ - tty_ldisc_halt(tty); tty_ldisc_reinit(tty); /* At this point we have a closed ldisc and we want to reopen it. We could defer this to the next open but