Re: 2.6.39-rc6-mmotm0506 - lockdep splat in RCU code on page fault

[Valdis . Kletnieks]

On Thu, 12 May 2011 02:47:05 PDT, "Paul E. McKenney" said:
> On Wed, May 11, 2011 at 07:11:34PM -0400, Valdis.Kletnieks@xxxxxx wrote:
> > My source has this:
> >
> >         raw_spin_lock_irqsave(&rnp->lock, flags);
> >         rnp->wakemask |= rdp->grpmask;
> >         invoke_rcu_node_kthread(rnp);
> >         raw_spin_unlock_irqrestore(&rnp->lock, flags);
> >
> > the last 2 lines swapped from what you diffed against.  I can easily work around
> > that, except it's unclear what the implications of the invoke_rcu moving outside
> > of the irq save/restore pair (or if it being inside is the actual root cause)...
>
> Odd...
>
> This looks to me like a recent -next -- I do not believe that straight
> mmotm has rcu_cpu_kthread_timer() in it.  The patch should apply to the
> last few days' -next kernels.

Ah. Found it. Your tree and current linux-next include this commit:

commit	1217ed1ba5c67393293dfb0f03c353b118dadeb4
tree	a765356c8418e134de85fd05d9fe6eda41de859c	tree | snapshot
parent	29ce831000081dd757d3116bf774aafffc4b6b20	commit | diff
rcu: permit rcu_read_unlock() to be called while holding runqueue locks

which includes this chunk:

@@ -1546,8 +1531,8 @@ static void rcu_cpu_kthread_timer(unsigned long arg)

        raw_spin_lock_irqsave(&rnp->lock, flags);
        rnp->wakemask |= rdp->grpmask;
-       invoke_rcu_node_kthread(rnp);
        raw_spin_unlock_irqrestore(&rnp->lock, flags);
+       invoke_rcu_node_kthread(rnp);
 }


but that was committed 4 days ago, and Andrew pulled linux-next for the -mmotm
6 days ago, so it's not in there.  The *rest* of your recent commits appear to
be in there though.  So that explains the patch failure to apply.

Attachment: pgp00000.pgp
Description: PGP signature