On Fri, May 29, 2009 at 7:18 PM, Scott Wood [off-list ref] wrote=
:

On Fri, May 29, 2009 at 12:56:13PM +0200, Frank Svendsb=F8e wrote:

quoted

FYI. The same applies to mpc8xx targets: No default host interrupt contr=

oller.

quoted

The following patch was needed for our target:
---

diff --git a/arch/powerpc/sysdev/mpc8xx_pic.c b/arch/powerpc/sysdev/mpc8=

xx_pic.c

quoted

index 5d2d552..92b2b66 100644

--- a/arch/powerpc/sysdev/mpc8xx_pic.c
+++ b/arch/powerpc/sysdev/mpc8xx_pic.c

@@ -186,6 +186,7 @@ int mpc8xx_pic_init(void)

=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 ret =3D -ENOMEM;
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 goto out;
=A0 =A0 =A0 =A0 }
+ =A0 =A0 =A0 =A0irq_set_default_host(mpc8xx_pic_host);
=A0 =A0 =A0 =A0 return 0;

This patch is whitespace mangled.

quoted

=A0out:
---
Maybe setting a default host ought to be mandatory? Or is doing the
mapping manually
(without device tree descriptions) considered being a hack?

I consider it a hack -- not so much doing it manually (though the device
tree is better), but relying on a default interrupt controller when doing
so. =A0IRQ numbers only make sense in the context of a specific
controller. =A0It's especially misleading on 8xx, which has separate
regular and CPM PICs.

-Scott

I agree, and was the reason I mentioned "hack". The patch wasn't meant
for commit,
just for reference (sorry for whitemangling ;-)

Regarding doing manual mapping: Is there another way to retrieve the
host controller
from a driver module without modifying kernel source? In case not, do you t=
hink
exporting the mpc8xx_pic_host symbol is a better solution?

Anyway, now that I'm beginning to understand dts I guess I might as
well just do it properly.

- Frank