On Thu, Sep 04, 2014 at 09:04:28AM -0700, Paul E. McKenney wrote:
On Thu, Sep 04, 2014 at 08:38:45PM +0800, Fengguang Wu wrote:
> On Tue, Sep 02, 2014 at 10:16:57AM -0700, Paul E. McKenney wrote:
> > On Tue, Sep 02, 2014 at 11:55:58AM -0500, Christoph Lameter wrote:
> > > On Tue, 2 Sep 2014, Paul E. McKenney wrote:
> > >
> > > > Added by ac1bea85781e (sched,rcu: Make cond_resched() report RCU
> > > > states), removed by 4a81e8328d379 (rcu: Reduce overhead of
> > > > checks for RCU). So, as you say, no effect on contemporary kernels.
> > >
> > > Well not sure what to make out of all of this....
> > Yep, still confused as to how the patch adding the definition could have
> > caused a failure. Fengguang, any thoughts?
> Yeah this is confusing.. I checked carefully and find that commit
> 0e98023 and afea227 are built on 2 quite different servers -- which
> might generate slightly different code. I'll fix this issue and make
> the build server selection more consistent.
Looking forward to seeing what shows up!
The new bisect catches commit 945fa9c631b04febe295a3a2a00c7e4a3cfb97db
("torture: Dump ftrace buffer when the RCU grace period stalls"). I
just reported it in another email.