[PATCH 2/3] switch frw to use local_soft_irq_pending

[PATCH 2/3] switch frw to use local_soft_irq_pending

[Christoph Hellwig]

The newly merged frv do_IRQ code calls softirq_pending(), but always with
the current cpu as argument - switch to local_softirq_pending().

Btw, this usage look bogus to me, any reason you need to call do_softirq
again after you did four lines above in irq_exit(), David?


--- 1.1/arch/frv/kernel/irq.c	2005-01-05 03:48:04 +01:00
+++ edited/arch/frv/kernel/irq.c	2005-01-07 12:27:31 +01:00
@@ -312,7 +312,7 @@
 
 	/* only process softirqs if we didn't interrupt another interrupt handler */
 	if ((__frame->psr & PSR_PIL) == PSR_PIL_0)
-		if (softirq_pending(cpu))
+		if (local_softirq_pending())
 			do_softirq();
 
 #ifdef CONFIG_PREEMPT

Re: [PATCH 2/3] switch frw to use local_soft_irq_pending

[David Howells]

Christoph Hellwig <hch@lst.de> wrote:

> The newly merged frv do_IRQ code calls softirq_pending(), but always with
> the current cpu as argument - switch to local_softirq_pending().
> 
> Btw, this usage look bogus to me, any reason you need to call do_softirq
> again after you did four lines above in irq_exit(), David?

Because irq_exit() doesn't call do_softirq() in 2.4 where I developed this
arch in the first place. And because 2.6 didn't malfunction due to
do_softirq() being called twice, I didn't notice.

Actually... it's probably also bad that it calls irq_exit() first and then
__clr_MASK(). That means it runs softirq processing with at least one
interrupt level disabled:-/

David