On Thursday, March 05, 2015 11:53:22 AM Catalin Marinas wrote:
On Wed, Mar 04, 2015 at 11:50:36PM +0100, Rafael J. Wysocki wrote:
On Wednesday, February 25, 2015 04:39:56 PM Hanjun Guo wrote:
[cut]
diff --git a/include/linux/acpi.h b/include/linux/acpi.h index c03d8d1..e27117a 100644 --- a/include/linux/acpi.h +++ b/include/linux/acpi.h @@ -557,6 +557,20 @@ static inline int acpi_device_modalias(struct device *dev, #endif /* !CONFIG_ACPI */ +#if defined(CONFIG_ACPI) && defined(CONFIG_ARM64) +static inline void acpi_irq_init(void) +{
- /*
* Hardcode ACPI IRQ chip initialization to GICv2 for now.* Proper irqchip infrastructure will be implemented along with* incoming GICv2m|GICv3|ITS bits.*/- acpi_gic_init();
+} +#else +static inline void acpi_irq_init(void) { } +#endif
I don't want this in a common header.
I don't like it either. What about adding it to a new header, linux/acpi_irq.h just for the dummy acpi_irq_init()? This would be similar to the DT equivalent, of_irq_init() in linux/of_irq.h and at some point it will gain more macros for declaring interrupt controllers in the ACPI context.
What we objected to previously was calling acpi_gic_init() directly from the core irqchip_init() and asked for something similar to the more generic of_irq_init(). The arm64-specific patch above is clearly a temporary hack until full support for multiple interrupt controllers is added (we asked for this several times in the past, but the ARM ACPI guys thought it's too much hassle ;). I don't fully get it since one of the platforms they target needs GICv2m support anyway).
Anyway, if we are to keep the temporary hack, I think we could define something like below (possibly in a new linux/acpi_irq.h which includes asm/irq.h):
#ifndef acpi_irq_init static inline void acpi_irq_init(void) { } #endif
And in the arm64 asm/irq.h:
static inline void acpi_irq_init(void) { /* * Hardcode ACPI IRQ chip initialization to GICv2 for now. * Proper irqchip infrastructure will be implemented along with * incoming GICv2m|GICv3|ITS bits. */ acpi_gic_init(); } #define acpi_irq_init acpi_irq_init
When the new infrastructure is in place, we can get rid of the #ifndef and arm64-specific acpi_irq_init().
Well, that sounds much better than the original patch.
Rafael