On 02/21, Jarek Poplawski wrote:
>
> > On Wed, 21 Feb 2007 01:19:41 +0300
> > Oleg Nesterov <[EMAIL PROTECTED]> wrote:
> >
> > > + p = container_of(work, struct net_bridge_port, carrier_check.work);
> > >
> > > rtnl_lock();
> > > - p = dev->br_port;
> > > - if (!p)
> > > - goto done;
>
On 02/20, Stephen Hemminger wrote:
>
> On Wed, 21 Feb 2007 01:19:41 +0300
> Oleg Nesterov <[EMAIL PROTECTED]> wrote:
>
> > static void release_nbp(struct kobject *kobj)
> > {
> > struct net_bridge_port *p
> > = container_of(kobj, struct net_bridge_port, kobj);
> > +
> > + dev_p
On Wed, Feb 21, 2007 at 09:23:45AM +0100, Jarek Poplawski wrote:
...
> I have known issues with RCU, but dare to disagree here.
> It's done during call_rcu, so anything RCU friendly shouldn't
> see this at the moment at all. It could be needed for those
> with refcounting - than it should be checke
On Tue, Feb 20, 2007 at 04:24:34PM -0800, Stephen Hemminger wrote:
> On Wed, 21 Feb 2007 01:19:41 +0300
> Oleg Nesterov <[EMAIL PROTECTED]> wrote:
>
> > If del_nbp()->cancel_delayed_work(carrier_check) fails, port_carrier_check()
> > may run later and access an already freed container (struct
> >
On Wed, 21 Feb 2007 01:19:41 +0300
Oleg Nesterov <[EMAIL PROTECTED]> wrote:
> If del_nbp()->cancel_delayed_work(carrier_check) fails, port_carrier_check()
> may run later and access an already freed container (struct net_bridge_port).
>
> With this patch, carrier_check owns a reference to "struct
If del_nbp()->cancel_delayed_work(carrier_check) fails, port_carrier_check()
may run later and access an already freed container (struct net_bridge_port).
With this patch, carrier_check owns a reference to "struct net_bridge_port",
not net_device, so it is always safe to acces the container.
port
