Re: [LOCKDEP BUG][2.6.36-rc1] xt_info_wrlock?

[Peter Zijlstra <peterz@infradead.org>]

On Mon, 2010-08-16 at 20:48 +0200, Eric Dumazet wrote:
> Le lundi 16 août 2010 à 20:36 +0200, Peter Zijlstra a écrit :
> > On Mon, 2010-08-16 at 14:16 -0400, Steven Rostedt wrote:
> > > @@ -729,8 +729,10 @@  static void get_counters(const struct
> > > xt_table_info *t,
> > >         local_bh_enable();
> > >         /* Processing counters from other cpus, we can let bottom half
> > > enabled,
> > >          * (preemption is disabled)
> > > +        * We must turn off lockdep to avoid a false positive.
> > >          */
> > >  
> > > +       lockdep_off();
> > >         for_each_possible_cpu(cpu) {
> > > 
> > nack!
> 
> 
> Interesting.
> 
> Care to elaborate ?

Adding lockdep_off() is just plain wrong, if you cannot describe the
locking there's a fair chance its wrong anyway.

As it stands there's only a single lockdep_off(), and that lives in NTFS
it looks like it could be annotated differently, but then, who cares
about NTFS anyway ;-)