Re: task_work_add/scheduler_tick: possible circular locking dependency detected

[Fengguang Wu <fengguang.wu@intel.com>]

On Mon, Oct 01, 2012 at 02:05:15PM +0200, Ingo Molnar wrote:
> 
> * Fengguang Wu <fengguang.wu@intel.com> wrote:
> 
> > On Mon, Oct 01, 2012 at 12:50:39PM +0200, Peter Zijlstra wrote:
> > > On Fri, 2012-09-28 at 19:34 +0800, Fengguang Wu wrote:
> > > > Peter,
> > > > 
> > > > I got the warning
> > > > 
> > > > [   10.412023] 
> > > > [   10.412611] ======================================================
> > > > [   10.413014] [ INFO: possible circular locking dependency detected ]
> > > > [   10.413014] 3.6.0-rc4-00098-g7eaffe9 #402 Not tainted
> > > > [   10.413014] -------------------------------------------------------
> > > > [   10.413014] init/1 is trying to acquire lock:
> > > > [   10.413014]  (&p->pi_lock){-.-.-.}, at: [<ffffffff81088214>] task_work_add+0x28/0x82
> > > > [   10.413014] 
> > > > [   10.413014] but task is already holding lock:
> > > > [   10.413014]  (&rq->lock){-.-.-.}, at: [<ffffffff8109c6ea>] scheduler_tick+0x3f/0xec
> > > > [   10.413014] 
> > > > [   10.413014] which lock already depends on the new lock.
> > > > [   10.413014] 
> > > > [   10.413014] 
> > > > 
> > > 
> > > The commit ac3d0da8f3290b3d394cdb7f50604424a7cd6092 should avoid this
> > > from happening, not sure what branch its on, but it was in tip before
> > > all this landed, so I guess its due to you testing sched/numa branch and
> > > not a merged branch like master or auto-next.
> > 
> > Peter, you are right, it's tested in tip:sched/numa. linux-next is
> > fine. Hmm, I should automatically test linux-next before raising the
> > problem, hehe.
> 
> tip:master is well-tested and generally a couple of days fresher 
> than linux-next, so in such a case where you are interested in 
> tip:sched/numa you should probably test tip:master.

OK, thanks for the tip!

Regards,
Fengguang