Re: [PATCH RFC] Revert "of/irq: Mark initialised interrupt controllers as populated"

Hello Rob,

On Sun, Aug 14, 2016 at 1:32 PM, Rob Herring [off-list ref] wrote:

On Sat, Aug 13, 2016 at 4:45 AM, Russell King
[off-list ref] wrote:

quoted

This reverts commit 15cc2ed6dcf91a8658e084be4e140147161819d7, which
causes a regression with iMX6 power domains.  iMX6 GPC contains both an
interrupt controller and power domains.  The iMX6 GPC code is setup to
register an interrupt controller using IRQCHIP_DECLARE(), but then to
register the power domains using the platform device.

This commit prevents the platform device being created, thereby breaking
iMX6 power domain support.

That commit also broke Exynos platforms since several devices are not
created (GIC, Exynos combiner and PMU system controller for
Exynos5420/5422/5800).

quoted

Signed-off-by: Russell King <redacted>
---
Please argue amongst yourselves about how to fix this regression caused
by this commit...

There's already a fix for this in my tree and -next. I will be sending
to Linus soon.

I guess you mean patch "[1/2] of/irq: Mark interrupt controllers as
populated before initialisation" [0] ?

Even with that patch on top of v4.8-rc2, things don't work for me. I
needed to revert setting the OF_POPULATED flag in of_irq_init for
things to work:

diff --git a/drivers/of/irq.c b/drivers/of/irq.c
index a2e68f740eda..5f9adc6c06e7 100644
--- a/drivers/of/irq.c
+++ b/drivers/of/irq.c

@@ -544,7 +544,7 @@ void __init of_irq_init(const struct of_device_id *matches)

                        list_del(&desc->list);

-                       of_node_set_flag(desc->dev, OF_POPULATED);
+                       //of_node_set_flag(desc->dev, OF_POPULATED);

                        pr_debug("of_irq_init: init %s (%p), parent %p\n",
                                 desc->dev->full_name,

I noticed that Philipp also sent a patch to clear that flag in
arch/arm/mach-imx/gpc.c [1]. I guess the same is needed in the drivers
for the devices that are not created for Exynos?

If that's the case, then I think that's more safe to revert Jon's
commit for now since is too risky for the -rc cycle and more platforms
may be affected.

[0]: https://patchwork.kernel.org/patch/9271361/
[1]: https://patchwork.kernel.org/patch/9271345/

Rob

Best regards,
Javier
--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html