Re: softlockup interaction with slow consoles

[Matt Mackall <mpm@selenic.com>]

On Tue, Feb 21, 2006 at 01:16:50AM -0800, David S. Miller wrote:
> From: Ingo Molnar <mingo@redhat.com>
> Date: Tue, 21 Feb 2006 04:09:58 -0500 (EST)
> 
> > i changed soft lockup detection to be turned off during bootup. That
> > should work around any boot-time warnings.
> 
> Excellent.

I don't like it. We should instead just have printk tickle the watchdog.
 
> > (if this can happen on a booted up system then the real fix would indeed
> > be to split up register_console()'s workload - that would also make it
> > more preemption-friendly. But at first sight it looks quite complex to
> > do.)
> 
> Agreed.  I thought about buffering in the slow console driver itself
> but that's bad because if it's a crash message we might not get the
> events (interrupts, or whatever) in order to make forward progress
> printing out the buffer, and thus we'd lose the valuable messages.

Absolutely.

Also note that sysrq-t can be substantially worse than the the
boot-time dump if you've got a lot of processes. So the boot-time hack
is insufficient. Even at 115kbps, this can be several seconds.

-- 
Mathematics is the supreme nostalgia of our time.