[Fwd: Re: -tip hangs my G5 during boot] -- Bug in "genirq: recordtrigger type"

From: Yinghai Lu
Date: Tue Oct 21 2008 - 03:29:55 EST




-------- Original Message --------
Subject: Re: -tip hangs my G5 during boot
Date: Fri, 10 Oct 2008 11:13:34 -0600
From: Chris Friesen <cfriesen@xxxxxxxxxx>
To: David Brownell <david-b@xxxxxxxxxxx>
CC: Ingo Molnar <mingo@xxxxxxx>, Thomas Gleixner <tglx@xxxxxxxxxxxxx>, Yinghai Lu <yinghai@xxxxxxxxxx>, "H. Peter Anvin" <hpa@xxxxxxxxx>
References: <48ECDFE2.1020609@xxxxxxxxxx> <200810091043.27149.david-b@xxxxxxxxxxx> <20081010153737.GC28977@xxxxxxx> <200810100958.38383.david-b@xxxxxxxxxxx>

David Brownell wrote:
> On Friday 10 October 2008, Ingo Molnar wrote:
>
>>>>> bisection gave the following as the culprit.
>>>>>
>>>>>commit 0c5d1eb77a8be917b638344a22afe1398236482b
>>>>>Author: David Brownell <dbrownell@xxxxxxxxxxxxxxxxxxxxx>
>>>>>
>>>>> genirq: record trigger type
>>
>>a quick review of the patch shows that it's about much more than just
>>recording the trigger type, it also changes the semantics of
>>set_irq_type():
>
>
> This is *exactly* what I meant when I wrote:
>
>
>>>>Other than that, I suspect something (new or old) depending on
>>>>either set_irq_type(irq, IRQ_TYPE_NONE) getting to an irq_chip
>>>>which doesn't treat it as a NOP, or ...
>
>
> As noted in my patch comments: as part of recording those flags,
> it made their semantics in set_irq_type() paths be identical to
> their semantics in request_irq() paths.
>
> There do seem to be some irq_chip implementations in the
> arch/powerpc/sysdev (and some Cell support) that treat
> IRQ_TYPE_NONE as other-than-NOP.

At least in my case, it was nothing so complicated. David's original
patch was passing "flags" rather than "type" down to __irq_set_trigger().
If I change it to pass "type" then everything works.

The following patch applies on top of -tip with David's original patch
re-applied.

Chris


From: Chris Friesen <cfriesen@xxxxxxxxxx>
Subject: [PATCH] fix set_irq_type() when recording trigger type

In set_irq_type() we want to pass the type rather than the current
interrupt state.

Signed-off-by: Chris Friesen <cfriesen@xxxxxxxxxx>


diff --git a/kernel/irq/chip.c b/kernel/irq/chip.c
index 570d1ea..1c178ae 100644
--- a/kernel/irq/chip.c
+++ b/kernel/irq/chip.c
@@ -132,7 +132,7 @@ int set_irq_type(unsigned int irq, unsigned int type)
return 0;

spin_lock_irqsave(&desc->lock, flags);
- ret = __irq_set_trigger(desc, irq, flags);
+ ret = __irq_set_trigger(desc, irq, type);
spin_unlock_irqrestore(&desc->lock, flags);
return ret;
}
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/