Re: Recent Padlock changes break ssh

On Thu, Jul 31, 2008 at 12:32:46PM +0200, Patrick Lamaizi?re wrote:
> Le Tue, 22 Jul 2008 10:14:49 +0200,
> Pawel Jakub Dawidek <pjd_at_FreeBSD.org> a écrit :
> 
> Hello,
> 
> > Could you try this patch? Those are the only changes that could
> > eventually change the behaviour.
> > 
> > 	http://people.freebsd.org/~pjd/patches/padlock.c.patch
> > 
> 
> I think that one problem is that the session id (ses->ses_id) is not
> updated when a free session is reused. The session id is set to zero by
> bzero() in padlock_freesession(). So we can have several active
> sessions with the same ses->ses_id == 0 if the sessions are reused.

Great catch! What do you think about using old sessid? I think it's ok
to do so and a bit safer, because session ID is only 32bit long so we
may get collision once we start from 0 again.

> padlock_freession()
>  	padlock_hash_free(ses);
>  	bzero(ses, sizeof(*ses));
>  	ses->ses_used = 0;
> 	TAILQ_INSERT_HEAD(&sc->sc_sessions, ses, ses_next);
> 
> and in padlock_newsession()
> 	/*
> 	 * Free sessions goes first, so if first session is used, we
> need to
> 	 * allocate one.
> 	 */
> 	ses = TAILQ_FIRST(&sc->sc_sessions);
> 	if (ses == NULL || ses->ses_used)
> 		ses = NULL;
> 	else {
> 		TAILQ_REMOVE(&sc->sc_sessions, ses, ses_next);
> 		ses->ses_used = 1;
> +		ses->ses_id = sc->sc_sid++;
> 		TAILQ_INSERT_TAIL(&sc->sc_sessions, ses, ses_next);

I'd replace 'sc->sc_sid++' with 'sid'.

-- 
Pawel Jakub Dawidek                       http://www.wheel.pl
pjd_at_FreeBSD.org                           http://www.FreeBSD.org
FreeBSD committer                         Am I Evil? Yes, I Am!