Re: [PATCH] printk: robustify printk

["Paul E. McKenney" <paulmck@linux.vnet.ibm.com>]

On Fri, Aug 08, 2008 at 12:14:28PM -0700, Andrew Morton wrote:
> On Fri, 08 Aug 2008 20:14:28 +0200
> Peter Zijlstra <a.p.zijlstra@chello.nl> wrote:
> 
> >  void wake_up_klogd(void)
> >  {
> > -	if (!oops_in_progress && waitqueue_active(&log_wait))
> > -		wake_up_interruptible(&log_wait);
> > +	unsigned long flags;
> > +	struct klogd_wakeup_state *kws;
> > +
> > +	if (!waitqueue_active(&log_wait))
> > +		return;
> > +
> > +	local_irq_save(flags);
> > +	kws = &__get_cpu_var(kws);
> > +	if (!kws->pending) {
> > +		kws->pending = 1;
> > +		call_rcu(&kws->head, __wake_up_klogd);
> > +	}
> > +	local_irq_restore(flags);
> >  }
> 
> Note that kernel/rcupreempt.c's flavour of call_rcu() takes
> RCU_DATA_ME().lock, so there are still code sites from which a printk
> can deadlock.  Only now, it is config-dependent.

The RCU callbacks are (and must be) invoked without holding any RCU
locks, so printk()s in RCU callbacks are perfectly permissible.  It would
not be -that- hard to eliminate the lock in call_rcu(), but yeccch...
It would be far easier to debug RCU without the benefit printk()s than
to deal with a preemptable RCU implementation that was based solely on
atomic instructions!!!

> >From a quick look it appears that large amounts of kernel/rcupreempt.c
> are now a printk-free zone.

The only printk() is in preemptable RCU's rcu_init() function, whose
only purpose is to mark the dmesg appropriately, and which should not
be a problem.

And that would be because I do a fair amount of prototyping and debugging
in user space, where one instead uses printf().  ;-)

							Thanx, Paul