Re: workqueue deadlock

[Andrew Morton <akpm@osdl.org>]

On Fri, 8 Dec 2006 08:23:01 +0530
Srivatsa Vaddagiri <vatsa@in.ibm.com> wrote:

> On Thu, Dec 07, 2006 at 11:37:00AM -0800, Andrew Morton wrote:
> > -static void flush_cpu_workqueue(struct cpu_workqueue_struct *cwq)
> > +/*
> > + * If cpu == -1 it's a single-threaded workqueue and the caller does not hold
> > + * workqueue_mutex
> > + */
> > +static void flush_cpu_workqueue(struct cpu_workqueue_struct *cwq, int cpu)
> 
> Lets say @cpu = 4
> 
> >  {
> >  	if (cwq->thread == current) {
> >  		/*
> >  		 * Probably keventd trying to flush its own queue. So simply run
> >  		 * it by hand rather than deadlocking.
> >  		 */
> > +		if (cpu != -1)
> > +			mutex_unlock(&workqueue_mutex);
> 
> Lets say we release the workqueue mutex here (events/4 is trying to
> flush its own workqueue). Immediately another CPU takes this mutex 
> (in CPU_DOWN_PREPARE) and brings down CPU4. In CPU_DEAD handling we now wait 
> on events/4 thread to exit (cleanup_workqueue_thread).
> 
> Couldnt this wait deadlock on :
> 
> >  		run_workqueue(cwq);
> 
> > +		if (cpu != -1)
> > +			mutex_lock(&workqueue_mutex);
> 
> events/4 thread itself wanting the same mutex above?
> 

Could do, not sure.  I'm planning on converting all the locking around here
to preempt_disable() though.