Re: printk: preempt_disable with long time resulting in softlockup/RCU stall issues

[Petr Mladek <pmladek@suse.com>]

On Wed 2015-07-08 14:00:17, Steven Rostedt wrote:
> On Wed, 08 Jul 2015 10:35:29 -0700
> Joe Perches <joe@perches.com> wrote:
> 
> > On Wed, 2015-07-08 at 19:04 +0200, Luis R. Rodriguez wrote:
> > > On Wed, Jul 08, 2015 at 05:08:35PM +0800, Gavin Hu wrote:
> > > > Hi,
> > > > 
> > > > Yes. We should disable the printk_limit feature when panic to avoid missing
> > > > messages.
> > > 
> > > Sounds like you have been looking into it and have a good idea of what you
> > > want to do, why not try it and send some RFC patches ?
> > > 
> > > While at it, then we could consider doing different things depending on the
> > > message type. KERN_EMERG would disable preemption, whereas KERN_INFO may not be
> > > so critical to require it.
> > 
> > That might be a bit difficult to implement with complete correctness
> > given KERN_EMERG use and continuation lines.
> > 
> 
> Or just have the current context determine what to do. If printk() was
> called with preemption or interrupts disabled, it flushes the full
> buffer before returning, otherwise it allows the writes to console be
> preempted.

I am afraid that any variant of this approach will make the original
problem even worse. There will be more messages waiting and it will
increase the chance of the softlockup.

I guess that people call printk() with disabled interrupts or
preemption with the naive hope that it will be fast (just store
the message in the ring buffer and come back).

I think that we really need to break the console output when it
takes too long and schedule it somewhere else.

Best Regards,
Petr