Re: [patch 2.6.28-rc2] at91_mci: workaround lockdep

[David Brownell <david-b@pacbell.net>]

On Tuesday 28 October 2008, Peter Zijlstra wrote:
> > > > Workaround: depend on LOCKDEP=n ... 
> > > 
> > > In all previous such cases it was deemed the IRQ handler should deal
> > > with whatever it gets.
> > 
> > In which case I'll wait until someone changes that IRQ handler (or that
> > ARM MM utility, or lockdep), and give up using AT91 platforms for sanity
> > testing kernel changes; lockdep is important, when it doesn't lie.
> > 
> > I do think that lockdep should warn when that it's ignoring such driver
> > requests, however.  I seem to have been tripping over it a lot lately,
> > and knowing that IRQ handlers were using strange modes would have saved
> > a bunch of time from being wasted.
> > 
> > Threaded IRQ handlers are going to need to rely even more on running
> > with IRQs enabled ... not to mention needing to sleep.  So it's clear
> > to me that there *are* lockdep issues yet to be addressed here.
> 
> Sure, care so send a patch fixing those? :-)

Here's one for the warning; that's the only one straightforward enough
to justify detouring from Real Work.  Plus, the IRQ threading patches
aren't that near a merge queue yet.  ;)

- Dave


========================== CUT HERE
From: David Brownell <dbrownell@users.sourceforge.net>

When lockdep turns on IRQF_DISABLED, emit a warning to make it
easier to track down problems this introduces in drivers that
expect handlers to run with IRQs enabled.

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
---
 kernel/irq/manage.c |    6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

--- a/kernel/irq/manage.c
+++ b/kernel/irq/manage.c
@@ -681,7 +681,11 @@ int request_irq(unsigned int irq, irq_ha
 	/*
 	 * Lockdep wants atomic interrupt handlers:
 	 */
-	irqflags |= IRQF_DISABLED;
+	if (!(irqflags & IRQF_DISABLED)) {
+		pr_warning("IRQ %d/%s: lockdep sets IRQF_DISABLED\n",
+				irq, devname);
+		irqflags |= IRQF_DISABLED;
+	}
 #endif
 	/*
 	 * Sanity-check: shared interrupts must pass in a real dev-ID,