Skip site navigation (1)Skip section navigation (2)
Date:      Wed, 11 May 2005 20:36:08 -0700 (PDT)
From:      Doug White <dwhite@gumbysoft.com>
To:        Ed Maste <emaste@phaedrus.sandvine.ca>
Cc:        stable@freebsd.org
Subject:   Re: Experimental ttwwakeup() panic patch
Message-ID:  <20050511202829.F11600@carver.gumbysoft.com>
In-Reply-To: <20050506145902.GA51724@sandvine.com>
References:  <20050503115344.S26250@carver.gumbysoft.com> <20050506145902.GA51724@sandvine.com>

next in thread | previous in thread | raw e-mail | index | archive | help
Sorry for the late reply on this.

On Fri, 6 May 2005, Ed Maste wrote:

> On Wed, May 04, 2005 at 08:54:23PM -0700, Doug White wrote:
>
> > > http://people.freebsd.org/~dwhite/tty.c.20050503.patch
> >
> > This patch has been committed and exists as rev 1.228.2.4 of
> > src/sys/kern/tty.c.  Please let me know if this fixes the panic for you,
> > or causes new problems :)
>
> For what it's worth, I've come across a similar crash during a test that
> repeatedly ssh'd into the machine.
>
> This was while opening a pty, not using serial console.  The symptoms
> seem similar; my tty struct has a struct knlist with a null struct mtx *.
> I haven't yet investigated in great detail or tried the patch.  I just
> wanted to offer this as another data point.
>
> Although kgdb gets confused during the backtrace (frames 7 and 8) the
> tf_eip is a cmpxchg in knote().

This is a problem mlaier and I may have fixed, at least against -CURRENT.
It appears that the process group alias in struct tty is accessed without
locking and its changing out from under us.

Give this a try:

http://people.freebsd.org/~mlaier/tty.t_pgrp.diff

It compiles but I haven't actually booted a kernel with it, so YMMV.

-- 
Doug White                    |  FreeBSD: The Power to Serve
dwhite@gumbysoft.com          |  www.FreeBSD.org



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20050511202829.F11600>