On Wed, 20 Oct 2021 16:14:07 +0100,
Florian Fainelli [off-list ref] wrote:

On 10/20/2021 1:24 AM, Marc Zyngier wrote:

quoted

On Tue, 19 Oct 2021 23:23:52 +0100,
Rob Herring [off-list ref] wrote:

quoted

  On Tue, Oct 19, 2021 at 4:43 PM Marc Zyngier [off-list ref] wrote:

quoted

diff --git a/include/linux/irqchip.h b/include/linux/irqchip.h
index ccf32758ea85..146a9d80a6a2 100644
--- a/include/linux/irqchip.h
+++ b/include/linux/irqchip.h

@@ -33,7 +33,15 @@ extern int platform_irqchip_probe(struct platform_device *pdev);
  #define IRQCHIP_PLATFORM_DRIVER_BEGIN(drv_name) \
  static const struct of_device_id drv_name##_irqchip_match_table[] = {

-#define IRQCHIP_MATCH(compat, fn) { .compatible = compat, .data = fn },
+/* Undefined on purpose */
+int typecheck_irq_init_cb(struct device_node *, struct device_node *,
+                         struct platform_device *);
+
+#define typecheck_irq_init_cb(fn)                                      \
+       __typecheck(typecheck_irq_init_cb, fn) ? fn : fn

That's nice! Shouldn't it also be used for IRQCHIP_DECLARE?

Absolutely. And enabling this shows that changing of_irq_init_cb_t
breaks *all users* of IRQCHIP_DECLARE(). Not an acceptable outcome
when we're at -rc5. >
Why can't the relevant drivers use of_find_device_by_node() instead?
That would allow us to keep the status-quo on of_irq_init_cb_t.

Rob had suggested several solutions, including using
of_find_device_by_node(), however updating of_irq_init_cb_t was
indicated to be the better way. I had intentionally not updated
IRQCHIP_DECLARE() because it would ignore the 3rd argument we passed
to it (platform_device *) so I thought.

In general, conflicting prototype always lead to the compiler
legitimately screwing something up, and you are left with a pile of
steaming crap to debug.

So *no* to that sort of trick.

If I am spinning a v6 using of_find_device_by_node() would that be
acceptable to you?

That'd be much better.

Thanks,

	M.

-- 
Without deviation from the norm, progress is not possible.