Re: lockdep: Warning & NULL ptr deref

[Peter Zijlstra <peterz@infradead.org>]

On Tue, 2012-02-28 at 16:23 +0200, Sasha Levin wrote:
> On Tue, Feb 28, 2012 at 4:20 PM, Peter Zijlstra <peterz@infradead.org> wrote:
> > On Tue, 2012-02-28 at 16:19 +0200, Sasha Levin wrote:
> >> On Tue, Feb 28, 2012 at 3:29 PM, Peter Zijlstra <peterz@infradead.org> wrote:
> >> > On Tue, 2012-02-28 at 10:24 -0500, Sasha Levin wrote:
> >> >>
> >> >> I'll send some patches to prevent the NULL deref itself which happens
> >> >> since a chunk of the code in lockdep.c assumed hlock_class() can't
> >> >> return NULL.
> >> >>
> >> > There's tons of code actually assuming that..
> >> >
> >> >> This won't fix the actual problem though, I'm not sure exactly how a
> >> >> held lock wouldn't have a class initialized to it.
> >> >
> >> > static inline struct lock_class *hlock_class(struct held_lock *hlock)
> >> > {
> >> >        if (!hlock->class_idx) {
> >> >                /*
> >> >                 * Someone passed in garbage, we give up.
> >> >                 */
> >> >                DEBUG_LOCKS_WARN_ON(1);
> >> >                return NULL;
> >> >        }
> >> >        return lock_classes + hlock->class_idx - 1;
> >> > }
> >> >
> >> > Typically that translates to severe memory corruption.
> >>
> >> I can also reproduce it easily by triggering sysrq once or twice after
> >> trinity has stopped.
> >
> > What's trinity ?
> 
> Dave Jones' fuzzer tool. The git tree is located here:
> http://codemonkey.org.uk/projects/trinity/trinity.git

Hmm, cute, does a kernel without lockdep explode too? If its proper
memory corruption one would expect that.