Re: [tip: locking/core] lockdep: Annotate irq_work

[Frederic Weisbecker <frederic@kernel.org>]

On Sat, Mar 21, 2020 at 07:12:49PM +0100, Sebastian Andrzej Siewior wrote:
> On 2020-03-21 17:40:58 [+0100], Frederic Weisbecker wrote:
> > > diff --git a/include/linux/irqflags.h b/include/linux/irqflags.h
> > > index 9c17f9c..f23f540 100644
> > > --- a/include/linux/irqflags.h
> > > +++ b/include/linux/irqflags.h
> > > @@ -69,6 +69,17 @@ do {						\
> > >  			current->irq_config = 0;	\
> > >  	  } while (0)
> > >  
> > > +# define lockdep_irq_work_enter(__work)					\
> > > +	  do {								\
> > > +		  if (!(atomic_read(&__work->flags) & IRQ_WORK_HARD_IRQ))\
> > > +			current->irq_config = 1;			\
> > 
> > So, irq_config == 1 means we are in a softirq? Are there other values for
> > irq_config? In which case there should be enums or something?
> > I can't find the patch that describes this.
> 
> 0 means as-is, 1 means threaded / sleeping locks are okay.

So that's the kind of comment we need :-)

Also how about current->irq_locking instead?

And something like:

enum {
    IRQ_LOCKING_NO_SLEEP,
    IRQ_LOCKING_CAN_SLEEP
}

> 
> > > --- a/kernel/time/tick-sched.c
> > > +++ b/kernel/time/tick-sched.c
> > > @@ -245,6 +245,7 @@ static void nohz_full_kick_func(struct irq_work *work)
> > >  
> > >  static DEFINE_PER_CPU(struct irq_work, nohz_full_kick_work) = {
> > >  	.func = nohz_full_kick_func,
> > > +	.flags = ATOMIC_INIT(IRQ_WORK_HARD_IRQ),
> > >  };
> > 
> > I get why these need to be in hardirq but some basic explanations for
> > ordinary mortals as to why those two specifically and not all the others
> > (and there are many) would have been nice.
> 
> Is the documentation patch in this series any good?

That describes the general rules but it doesn't tell anything about the
details of this patch. Especially why RCU and nohz_full irq works in particular
are special here and why it's fine for others to execute in softirq.

Thanks.

> > Thanks.
> 
> Sebastian