Re: [PATCH 1/3] notifier: Fix broken error handling pattern

[Peter Zijlstra <peterz@infradead.org>]

On Mon, Jun 24, 2019 at 05:21:07PM -0500, Josh Poimboeuf wrote:
> On Mon, Jun 24, 2019 at 11:18:44AM +0200, Peter Zijlstra wrote:
> > The current notifiers have the following error handling pattern all
> > over the place:
> > 
> > 	int nr;
> > 
> > 	ret = __foo_notifier_call_chain(&chain, val_up, v, -1, &nr);
> > 	if (err & NOTIFIER_STOP_MASK)
> 
> s/err/ret/

-ETOOWARM :-)

> > 		__foo_notifier_call_chain(&chain, val_down, v, nr-1, NULL)
> > 
> > And aside from the endless repetition thereof, it is broken. Consider
> > blocking notifiers; both calls take and drop the rwsem, this means
> > that the notifier list can change in between the two calls, making @nr
> > meaningless.
> > 
> > Fix this by replacing all the __foo_notifier_call_chain() functions
> > with foo_notifier_call_chain_error() that embeds the above patter, but
> > ensures it is inside a single lock region.
> 
> The name "notifier_call_chain_error()" seems confusing, it almost sounds
> like it's notifying an error code.  Then again, I can't really think of
> a more reasonably succinct name.

I;m not attached to the name; I very much ran out of ideas and just
typed something.

> > @@ -25,8 +25,23 @@ static int cpu_pm_notify(enum cpu_pm_eve
> >  	 * RCU know this.
> >  	 */
> >  	rcu_irq_enter_irqson();
> > -	ret = __atomic_notifier_call_chain(&cpu_pm_notifier_chain, event, NULL,
> > -		nr_to_call, nr_calls);
> > +	ret = atomic_notifier_call_chain(&cpu_pm_notifier_chain, event, NULL);
> > +	rcu_irq_exit_irqson();
> > +
> > +	return notifier_to_errno(ret);
> > +}
> > +
> > +static int cpu_pm_notify_error(enum cpu_pm_event event_up, enum cpu_pm_event event_down)
> > +{
> > +	int ret;
> > +
> > +	/*
> > +	 * __atomic_notifier_call_chain has a RCU read critical section, which
> 
> __atomic_notifier_call_chain() no longer exists.
> 
> > +	 * could be disfunctional in cpu idle. Copy RCU_NONIDLE code to let
> 
> "dysfunctional"

That's copy paste, I don't think I've read the comment, my bad.

> > @@ -156,43 +169,30 @@ int atomic_notifier_chain_unregister(str
> >  }
> >  EXPORT_SYMBOL_GPL(atomic_notifier_chain_unregister);
> >  
> > -/**
> > - *	__atomic_notifier_call_chain - Call functions in an atomic notifier chain
> > - *	@nh: Pointer to head of the atomic notifier chain
> > - *	@val: Value passed unmodified to notifier function
> > - *	@v: Pointer passed unmodified to notifier function
> > - *	@nr_to_call: See the comment for notifier_call_chain.
> > - *	@nr_calls: See the comment for notifier_call_chain.
> > - *
> > - *	Calls each function in a notifier chain in turn.  The functions
> > - *	run in an atomic context, so they must not block.
> > - *	This routine uses RCU to synchronize with changes to the chain.
> > - *
> > - *	If the return value of the notifier can be and'ed
> > - *	with %NOTIFY_STOP_MASK then atomic_notifier_call_chain()
> > - *	will return immediately, with the return value of
> > - *	the notifier function which halted execution.
> > - *	Otherwise the return value is the return value
> > - *	of the last notifier function called.
> > - */
> 
> Why remove the useful comment?

Because I delete the whole function ?