Re: 4.14 backport request for dbdda842fe96f: "printk: Add console owner and waiter logic to load balance console writes"

[Sergey Senozhatsky]

On (10/21/18 11:09), Daniel Wang wrote:
> 
> Just got back from vacation. Thanks for the continued discussion. Just so
> I understand the current state. Looks like we've got a pretty good explanation
> of what's going on (though not completely sure), and backporting Steven's
> patches is still the way to go?

Up to -stable maintainers.

Note, with or without Steven's patch, the non-reentrable consoles are
still non-reentrable, so the deadlock is still there:

	spin_lock_irqsave(&port->lock, flags)
	 <NMI>
	  panic()
	   console_flush_on_panic()
	    spin_lock_irqsave(&port->lock, flags)		// deadlock


// And I wouldn't mind to have more reviews/testing on [1].


Another deadlock scenario could be the following one:

	printk()
	 console_trylock()
	  down_trylock()
	   raw_spin_lock_irqsave(&sem->lock, flags)
	    <NMI>
	     panic()
	      console_flush_on_panic()
	       console_trylock()
	        raw_spin_lock_irqsave(&sem->lock, flags)	// deadlock

There are no patches addressing this one at the moment. And it's
unclear if you are hitting this scenario.


> I see that Sergey had sent an RFC series for similar things. Are those
> trying to solve the deadlock problem in a different way?

Umm, I wouldn't call it "another way". It turns non-reentrant serial
consoles to re-entrable ones. Did you test patch [1] from that series
on you environment, by the way?

[1] lkml.kernel.org/r/20181016050428.17966-2-sergey.senozhatsky@xxxxxxxxx

	-ss