Re: + panic-avoid-the-extra-noise-dmesg.patch added to -mm tree

[Petr Mladek]

On Tue 2018-12-11 17:07:43, Sergey Senozhatsky wrote:
> On (12/10/18 16:57), Petr Mladek wrote:
> > > > (masked out) and on panic_cpu disables only SDEI (interrupts from firmware,
> > > > if I got it right); so it seems that arm64 can handle IRQs after panic. And
> > > > if there are platforms that handle IRQ (including sysrq) after panic, then
> > > > both options - making printk a noop or keeping local irqs off - maybe can
> > > > cause some problems. Or maybe not. We better ask arch people.
> > > 
> > > Yes, this is very valid concern. And after Petr and you raised it, I did
> > > some experiments with 3 x86 platforms at my hand, one Apollolake IOT device
> > > with serial console, one IvyBridge laptop and one Kabylake NUC, the magic key
> > > all works well before panic, and fails after panic. But I did remember the
> > > PageUp/PageDown key worked on some laptop years ago. And you actually raised a
> > > good question: what do we expect for the post-panic kernel?
> > 
> > I am not sure why it does not work. But it would be nice if sysrq
> > worked.
> 
> Absolutely.
> 
> [..]
> > I still think that calming down printk() is acceptable when
> > it can be restored from sysrq.
> 
> I would agree; peeking one of the two solutions, printk patch is
> probably preferable.
> 
> > I think that only few people might be interested into debugging
> > post-panic problems. We could print a warning for them about
> > that printk() has got disabled.
> 
> Dunno. This _maybe_ (speculation!) can upset folks on those platforms
> that have sysrq working after panic. printk is a common code.
> 
> I'm probably missing a lot of things here, but just in case, I'm not
> sure at which point the idea of patching some files under arch/x86
> directory was ruled out and why.

I suggested to clear the panic_blinking (or whatever name) in
__handle_sysrq(). The idea is that sysrq needs manual intervention.
It allows to see the original message before it gets overridden
by a potential sysrq-related output.

It assumes that sysrq is the only interesting operation when
printk() might be useful at this state.

Best Regards,
Petr