On Thu, Sep 20, 2012 at 10:47:25PM +0000, Paul Walmsley wrote:
> On Thu, 20 Sep 2012, Paul E. McKenney wrote:
>
> > On Thu, Sep 20, 2012 at 09:49:13PM +0000, Bruce, Becky wrote:
> >
> > > OK, people, you can stop heckling me about "sent from my iPhone" - I'm in
> > > the wilds of rural Louisiana with really bad internet service and was
> > > trying to work on my phone (but, alas, did not notice the CC list
> > > included the entire universe). Shame on me.
> > >
> > > With the above set, I don't seem to see any stalls with the RCU timeout
> > > set to 60s (the default). I left the board running for 25 minutes; I
> > > will fire it up again later and let it run for a bit longer, but usually
> > > I end up seeing the problem pretty quickly so I don't expect that to
> > > result in anything. I also didn't see any stalls on Paul's RCU tree as
> > > of a week ago at 60s, so as far as I can tell the CPU_IDLE stuff didn't
> > > have any impact (it wasn't on when I tested Paul's tree).
> > >
> > > If I drop the timeout to 5s as Paul M. suggested for debug a while back,
> > > I do see stalls (both with CPU_IDLE stuff and without).
> > >
> > > I'm using the default omap2plus config, with RCU stall info enabled and
> > > the cpu idle stuff turned on (console dump below). This is a Panda ES
> > > 1.1 (OMAP4460)
> >
> > Thank you for the testing, Becky!
> >
> > Paul Walmsley, please let me know if the config below doesn't clear things
> > up for you or if there is some reason why this config is infeasible.
>
> Will certainly test it here once things settle down. But I wouldn't
> consider the use of CONFIG_CPU_IDLE to be a 'fix' until the problem can be
> understood. We use !CONFIG_CPU_IDLE kernels for testing here.
Fair point. I am wondering whether there is some path into the idle
loop that somehow avoids telling RCU that the CPU has in face entered
idle. There needs to be an rcu_idle_enter() call on the way to idle,
otherwise RCU CPU stall warnings are expected behavior.
Thanx, Paul
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
