Re: [PATCH] printk: Ratelimit messages printed by console drivers

[Petr Mladek]

On Fri 2018-04-20 11:13:07, Steven Rostedt wrote:
> On Fri, 20 Apr 2018 16:57:20 +0200
> Petr Mladek <pmladek@xxxxxxxx> wrote:
> 
> 
> > No, call_console_drivers() is done with interrupts disabled:
> > 
> > 		console_lock_spinning_enable();
> > 
> > 		stop_critical_timings();	/* don't trace print latency */
> >  ---->		call_console_drivers(ext_text, ext_len, text, len);  
> > 		start_critical_timings();
> > 
> > 		if (console_lock_spinning_disable_and_check()) {
> >  ---->			printk_safe_exit_irqrestore(flags);  
> > 			goto out;
> > 		}
> > 
> >  ---->		printk_safe_exit_irqrestore(flags);  
> > 
> > They were called with interrupts disabled for ages, long before
> > printk_safe. In fact, it was all the time in the git kernel history.
> > 
> > Therefore only NMIs are in the game. And they should be solved
> > by the above change.
> > 
> 
> Really?
> 
> 
>   console_trylock_spinning(); /* console_owner now equals current */

No, console_trylock_spinning() does not modify console_owner. The
handshake is done using console_waiter variable.

console_owner is really set only between:

    console_lock_spinning_enable()
    console_lock_spinning_disable_and_check()

and this entire section is called with interrupts disabled.

Best Regards,
Petr