lkml.org 
[lkml]   [2009]   [Mar]   [13]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [tip:core/ipi] generic-ipi: eliminate spurious pointlessWARN_ON()s

* Peter Zijlstra <peterz@infradead.org> wrote:

> On Fri, 2009-03-13 at 09:21 +0000, Jan Beulich wrote:
> > >>> Peter Zijlstra <peterz@infradead.org> 13.03.09 09:54 >>>
> > >Wouldn't leaving them in place but changing them to:
> > >
> > >WARN_ON(irqs_disabled() && system_state == SYSTEM_RUNNING);
> > >
> > >be clearer?
> >
> > I don't think that would be precise: system_state gets set to
> > SYSTEM_RUNNING much later than APs get brought up (i.e. there are
> > cases where the WARN_ON()s could validly trigger with SYSTEM_BOOTING),
> > and also doesn't cover states > SYSTEM_RUNNING (where, again, after
> > perhaps having brought down all APs the warnings could become pointless,
> > but the warnings could be meaningful as long as there are still some APs
> > online).
> >
> > While from an abstract code reading perspective your suggestion might
> > seem reasonable, the changed placement really reflects what the warning
> > is trying to warn about - a potential deadlock which cannot occur under
> > the conditions filtered out by conditionals the warnings were moved
> > beyond.
>
> How about?
>
> WARN_ON_ONCE(irqs_disabled() && !oops_in_progress)

ok, that's indeed better - i've done that - see the upcoming
commit messages in this thread.

I also changed all of panic() to be covered by oops_in_progress
- see the other commit. Eric: that should make kexec [a tiny
bit] more robust too, agreed?

Ingo


\
 
 \ /
  Last update: 2009-03-13 11:41    [from the cache]
©2003-2011 Jasper Spaans