Message-ID: <1474464010.32518.3.camel@redhat.com>
Subject: Re: [PATCH 1/2] NFSD: notifiers registration cleanup
From: Jeff Layton <jlayton@redhat.com>
To: Vasily Averin <vvs@virtuozzo.com>, linux-kernel@vger.kernel.org,
        linux-nfs@vger.kernel.org
Cc: "J. Bruce Fields" <bfields@fieldses.org>,
        Scott Mayhew <smayhew@redhat.com>
Date: Wed, 21 Sep 2016 09:20:10 -0400
In-Reply-To: <6bce434f-4798-42bf-8d4f-e1f33ceb8ad6@virtuozzo.com>
References: <6bce434f-4798-42bf-8d4f-e1f33ceb8ad6@virtuozzo.com>
Content-Type: text/plain; charset="UTF-8"
Mime-Version: 1.0
Sender: linux-nfs-owner@vger.kernel.org

On Wed, 2016-09-21 at 15:33 +0300, Vasily Averin wrote:
> By design notifier can be registered once only,
> however nfsd registers the same inetaddr notifiers per net-namespace.
> When this happen it corrupts list of notifiers,
> as result some notifiers can be not called on proper event,
> traverse on list can be cycled forever,
> and second unregister can access already freed memory.
> 
> fixes: 36684996 ("nfsd: Register callbacks on the inetaddr_chain and inet6addr_chain")
> > Signed-off-by: Vasily Averin <vvs@virtuozzo.com>
> ---
>  fs/nfsd/nfssvc.c | 17 +++++++++++++----
>  1 file changed, 13 insertions(+), 4 deletions(-)
> 
> diff --git a/fs/nfsd/nfssvc.c b/fs/nfsd/nfssvc.c
> index 45007ac..7f8914f 100644
> --- a/fs/nfsd/nfssvc.c
> +++ b/fs/nfsd/nfssvc.c
> @@ -366,14 +366,20 @@ static struct notifier_block nfsd_inet6addr_notifier = {
>  };
>  #endif
>  
> +static atomic_t nfsd_notifier_refcount = ATOMIC_INIT(0);
> +
>  static void nfsd_last_thread(struct svc_serv *serv, struct net *net)
>  {
> >  	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
>  
> > -	unregister_inetaddr_notifier(&nfsd_inetaddr_notifier);
> > +	/* check if the notifier still has clients */
> > +	if (atomic_dec_return(&nfsd_notifier_refcount) == 0) {
> > +		unregister_inetaddr_notifier(&nfsd_inetaddr_notifier);
>  #if IS_ENABLED(CONFIG_IPV6)
> > -	unregister_inet6addr_notifier(&nfsd_inet6addr_notifier);
> > +		unregister_inet6addr_notifier(&nfsd_inet6addr_notifier);
>  #endif
> > +	}
> +
> >  	/*
> >  	 * write_ports can create the server without actually starting
> >  	 * any threads--if we get shut down before any threads are
> @@ -488,10 +494,13 @@ int nfsd_create_serv(struct net *net)
> >  	}
>  
> >  	set_max_drc();
> > -	register_inetaddr_notifier(&nfsd_inetaddr_notifier);
> > +	/* check if the notifier is already set */
> > +	if (atomic_inc_return(&nfsd_notifier_refcount) == 1) {
> > +		register_inetaddr_notifier(&nfsd_inetaddr_notifier);
>  #if IS_ENABLED(CONFIG_IPV6)
> > -	register_inet6addr_notifier(&nfsd_inet6addr_notifier);
> > +		register_inet6addr_notifier(&nfsd_inet6addr_notifier);
>  #endif
> > +	}
> > >  	do_gettimeofday(&nn->nfssvc_boot);		/* record boot time */
> >  	return 0;
>  }

Good catch. I'm not very fond of the refcounting this here but it
should
serve the purpose and I don't have anything better to suggest. FWIW, I
think the nfsd_mutex is held during all of these operations so we
probably don't need atomics for the refcount.

--
Jeff Layton <jlayton@redhat.com>