Date: Tue, 1 Apr 2014 11:32:45 -0700
From: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
To: Dave Jones <davej@redhat.com>, Linux Kernel <linux-kernel@vger.kernel.org>
Subject: Re: rcu_prempt stalls / lockup
Message-ID: <20140401183245.GA12473@linux.vnet.ibm.com>
Reply-To: paulmck@linux.vnet.ibm.com
References: <20140331230241.GA30019@redhat.com>
 <20140331232220.GP4284@linux.vnet.ibm.com>
 <20140331233552.GB30019@redhat.com>
 <20140401004801.GQ4284@linux.vnet.ibm.com>
 <20140401150849.GA14757@redhat.com>
 <20140401153032.GT4284@linux.vnet.ibm.com>
 <20140401172244.GA10363@redhat.com>
 <20140401175545.GV4284@linux.vnet.ibm.com>
 <20140401180414.GA12326@redhat.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20140401180414.GA12326@redhat.com>
User-Agent: Mutt/1.5.21 (2010-09-15)
Sender: linux-kernel-owner@vger.kernel.org

On Tue, Apr 01, 2014 at 02:04:14PM -0400, Dave Jones wrote:
> On Tue, Apr 01, 2014 at 10:55:45AM -0700, Paul E. McKenney wrote:
>  > >  > > so kernel space still works like before, but userspace is locked up.
>  > >  > 
>  > >  > Interesting.  I suspect that if you reverted the rest of this merge
>  > >  > window's RCU patches, you would get the same result.
> 
> Something that occurred to me is that this might be something in the x86 merge
> that's just changing timings enough to expose this problem.
> At some point this evening, I'll try bisecting it if we don't get any closer.

OK.  ;-)

>  > > [ 1953.672735] INFO: Stall ended before state dump start, gp_kthread state: 0x2
>  > > [ 2148.608132] INFO: rcu_preempt detected stalls on CPUs/tasks:
>  > > [ 2148.609140] 	(detected by 0, t=104027 jiffies, g=47728, c=47727, q=0)
>  > > etc etc.
>  > 
>  > Waiting uninterruptibly.  Presumably blocked on mutex_lock().  But
>  > you have CONFIG_PROVE_LOCKING(), so any deadlocks should have been
>  > reported.
> 
> Lockdep had reported something a little earlier (timestamped at 1108.xxxxxx)
> but that's a known false-positive in xfs.

Yep, I would be very surprised if that was related to the grace-period hang.

>  > Given that you have CONFIG_RCU_TRACE=y, could you please enable the
>  > following trace events and dump the trace before things hang?
>  > 
>  > 	trace_event=rcu:rcu_grace_period,rcu:rcu_grace_period_init
>  > 
>  > If it is not feasible to dump the trace before things hang, let me
>  > know, and I will work out some other diagnostic regime.
> 
> I'll give that a shot when I get back in a few hours.

Cool!

							Thanx, Paul

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/