Prev: drivers/net: remove network drivers' last few uses of IRQF_SAMPLE_RANDOM
Next: ALSA: Add initial support for Thinkpad T410s HDA codec.
From: David Miller on 15 Apr 2010 03:00
From: Ingo Molnar <mingo(a)elte.hu>
Date: Thu, 15 Apr 2010 08:49:40 +0200

> Btw., WARN_ON trapping on PowerPC is clearly a PowerPC bug - there's a good
> reason we have WARN_ON versus BUG_ON - it should be fixed.

I disagree, an implementation should be allowed to use the most
efficient implementation possible for both interfaces.

I would be using traps for both on sparc64 if that were really
feasible on sparc64 (and actually with gcc-4.5's "asm goto" it might
actually be now)

The WARN and BUG macros, when implemented without traps, have serious
implications for overall code size and register pressure.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
From: Stephen Rothwell on 15 Apr 2010 06:10
Hi Ingo,

On Thu, 15 Apr 2010 08:49:40 +0200 Ingo Molnar <mingo(a)elte.hu> wrote:
>
> Ok, we'll fix the warning.

Thanks.

--
Cheers,
Stephen Rothwell sfr(a)canb.auug.org.au
http://www.canb.auug.org.au/~sfr/
From: Benjamin Herrenschmidt on 15 Apr 2010 22:00
On Thu, 2010-04-15 at 09:32 +0200, Ingo Molnar wrote:
> It trades robustness for slightly better space/code efficiency.
>
> Such a trap based mechanism exists on x86 as well and we use it for BUG_ON().
> We intentionally dont use it to generate warnings and dont override __WARN(),
> because it would blow up way too often when a warning triggers in some
> sensitive codepath that cannot take a trap.
>
> Anyway, the warning obviously has to be fixed - but the boot crash itself is
> PowerPC's own doing.

Well, yes and no, as I explained in a separate branch of that thread. We
indeed can't cope with a WARN in that spot because it goes recursive.

Now the reason we have this double-enable is due afaik to the way I
implemented IRQ trace, because things like syscalls basically
force-enable IRQs on powerpc and I don't necessarily have tracking
informations in the exception return path of what the "old" value was.

I need to double check what the exact scenario here is and whether I can
fix it but it's one of those cases where what lockdep is warning about
isn't actually an error I believe.

Cheers,
Ben.

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
From: Peter Zijlstra on 16 Apr 2010 06:40
On Thu, 2010-04-15 at 19:15 +0200, Frederic Weisbecker wrote:
> > that looks rather ugly. Why not do a raw:
> >
> > this_cpu_inc(lockdep_stats.redundant_hardirqs_on);
> >
> > which basically open-codes debug_atomic_inc(), but without the warning?
>
>
> Because that would open a race against interrupts that might
> touch lockdep_stats.redundant_hardirqs_on too.


How so, its a pure per-cpu variable right? so either the increment
happens before the interrupts hits, or after, in either case there
should not be a race with interrupts.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
 | 
Pages: 1
Prev: drivers/net: remove network drivers' last few uses of IRQF_SAMPLE_RANDOM
Next: ALSA: Add initial support for Thinkpad T410s HDA codec.