Re: [patch] fix netconsole hang with alt-sysrq-t

[Matt Mackall <mpm@selenic.com>]

On Fri, Aug 13, 2004 at 12:39:36AM +0300, Muli Ben-Yehuda wrote:
> On Thu, Aug 12, 2004 at 05:32:21PM -0400, Jeff Moyer wrote:
> > ==> Regarding Re: [patch] fix netconsole hang with alt-sysrq-t; Muli Ben-Yehuda <mulix@mulix.org> adds:
> > 
> > mulix> On Thu, Aug 12, 2004 at 05:01:18PM -0400, Jeff Moyer wrote:
> > >> So how do you want to deal with this case?  We could do something like:
> > >> 
> > >> int cpu = smp_processor_id();
> > 
> > mulix> That doesn't look right, unless I'm missing something, you could get
> > mulix> preempted here (between the smp_processor_id() and the
> > mulix> local_irq_save() and end up with 'cpu' pointing to the wrong CPU.
> > 
> > Would a preempt_disable() be too hideous?  Other suggestions?
> 
> Maybe, but we could hide it in get_cpu() / put_cpu() ;-)

Yes, let's. I'll have to think about this general approach a bit more though.

-- 
Mathematics is the supreme nostalgia of our time.